Tech Guides - Data

On November 9, 2015, a storm loomed over the SF Bay area creating major outages. At Mountain View, California, Google engineers were busy creating a storm of their own. That day, Sundar Pichai announced to the world that TensorFlow, their machine learning system, was going Open Source. He said: “...today we’re also open-sourcing TensorFlow. We hope this will let the machine learning community—everyone from academic researchers, to engineers, to hobbyists—exchange ideas much more quickly, through working code rather than just research papers.” That day the tech world may not have fully grasped the gravity of the announcement but those in the know knew it was a pivotal moment in Google’s transformational journey into an AI first world. How did TensorFlow begin? TensorFlow was part of a former Google product called DistBelief. DistBelief was responsible for a program called DeepDream. The program was built for scientists and engineers to visualise how deep neural networks process images. As fate would have it, the algorithm went viral and everyone started visualising abstract and psychedelic art in it. Although people were having fun playing with image forms, they were unaware of the technology that powered those images - neural networks and deep learning - the exact reason why TensorFlow was built for. TensorFlow is a machine learning platform that allows one to run a wide range of algorithms like the aforementioned neural networks and deep learning based projects. TensorFlow with its flexibility, high performance, portability, and production-readiness is changing the landscape of artificial intelligence and machine learning. Be it face recognition, music, and art creation or detecting clickbait headline for blogs, the use cases are immense. With Google open sourcing TensorFlow, the platform that powers Google search and other smart Google products is now accessible to everyone - researchers, scientists, machine learning experts, students, and others. So why did Google open source TensorFlow? Yes, Google made a world of difference to the machine learning community at large by open sourcing TensorFlow. But what was in it for Google? As it turns out, a whole lot. Let’s look at a few. Google is feeling the heat from rival deep learning frameworks Major deep learning frameworks like Theano, Keras, etc., were already open source. Keeping a framework proprietary was becoming a strategic disadvantage as most DL core users i.e. scientists, engineers, and academicians prefer using open source software for their work. “Pure” researchers and aspiring “Phds” are key groups that file major patents in the world of AI. By open sourcing TensorFlow, Google gave this community access to a platform it backs to power their research. This makes migrating the world’s algorithms from other deep learning tools onto TensorFlow theoretically possible. AI as a trend is clearly here to stay and Google wants a platform that leads this trend. An open source TensorFlow can better support the Google Brain project Behind all the PR, Google does not speak much about its pet project Google Brain. When Sundar Pichai talks of Google’s transformation from Search to AI, this project is doing all the work behind the scenes. Google Brain is headed by some of the best minds in the industry like Jeff Dean, Geoffery Hilton, Andrew NG, among many others. They developed TensorFlow and they might still have some state-of-the-art features up their sleeves privy only to them. After all, they have done a plethora of stunning research in areas like parallel computing, machine intelligence, natural language processing and many more. With TensorFlow now open sourced, this team can accelerate the development of the platform and also make significant inroads into areas they are currently researching on. This research can then potentially develop into future products for Google which will allow them to expand their AI and Cloud clout, especially in the enterprise market. Tapping into the collective wisdom of the academic intelligentsia Most innovations and breakthroughs come from universities before they go mainstream and become major products in enterprises. AI, still making this transition, will need a lot of investment in research. To work on difficult algorithms, researchers will need access to sophisticated ML frameworks. Selling TensorFlow to universities is an old school way to solve the problem - that’s why we no longer hear about products like LabView. Instead, by open-sourcing TensorFlow, the team at Google now has the world’s best minds working on difficult AI problems on their platform for free. As these researchers start writing papers on AI using TensorFlow, it will keep adding to the existing body of knowledge. They will have all the access to bleeding-edge algorithms that are not yet available in the market. Their engineers could simply pick and choose what they like and start developing commercially ready services. Google wants to develop TensorFlow as a platform-as-a-service for AI application development An advantage of open-sourcing a tool is that it accelerates time to build and test through collaborative app development. This means most of the basic infrastructure and modules to build a variety of TensorFlow based applications will already exist on the platform. TensorFlow developers can develop and ship interesting modular products by mixing and matching code and providing a further layer of customization or abstraction. What Amazon did for storage with AWS, Google can do for AI with TensorFlow. It won’t come as a surprise if Google came up with their own integrated AI ecosystem with TensorFlow on the Google Cloud promising you the AI resources your company would need. Suppose you want a voice based search function on your ecommerce mobile application. Instead, of completely reinventing the wheel, you could buy TensorFlow powered services provided by Google. With easy APIs, you can get voice based search and save substantial developer cost and time. Open sourcing TensorFlow will help Google to extend their talent pipeline in a competitive Silicon Valley jobs market Hiring for AI development is  competitive in the Silicon Valley as all major companies vie for attention from the same niche talent pool. With TensorFlow made freely available, Google’s HR team can quickly reach out to a talent pool specifically well versed with the technology and also save on training cost. Just look at the interest TensorFlow has generated on a forum like StackOverflow: This indicates that growing number of users are asking and inquiring about TensorFlow. Some of these users will migrate into power users who the Google HR team can tap into. A developer pool at this scale would never have been possible with a proprietary tool. Replicating the success and learning from Android Agreed, a direct comparison with Android is not possible. However, the size of the mobile market and Google’s strategic goal of mobile-first when they introduced Android bear striking similarity with the nascent AI ecosystem we have today and Google’s current AI-first rhetoric. In just a decade since its launch, Android now owns more than 85% of the smartphone mobile OS market. Piggybacking on Android’s success, Google now has control of mobile search (96.19%), services (Google Play), a strong connection with the mobile developer community and even a viable entry into the mobile hardware market. Open sourcing Android did not stop Google from making money. Google was able to monetize through other ways like mobile search, mobile advertisements, Google Play, devices like Nexus, mobile payments, etc. Google did not have all this infrastructure planned and ready before Android was open sourced - It innovated, improvised, and created along the way. In the future, we can expect Google to adopt key learnings from its Android growth story and apply to TensorFlow’s market expansion strategy. We can also see supporting infrastructures and models for commercialising TensorFlow emerge for enterprise developers. [dropcap]T[/dropcap]he road to AI world domination for Google is on the back of an open sourced TensorFlow platform. It appears not just exciting but also promises to be one full of exponential growth, crowdsourced innovation and learnings drawn from other highly successful Google products and services. The storm that started two years ago is surely morphing into a hurricane. As Professor Michael Guerzhoy of University of Toronto quotes in Business Insider “Ten years ago, it took me months to do something that for my students takes a few days with TensorFlow.”

The answer is yes, and no. This is a question that could've easily been applied to textile manufacturing in the 1890s, and could've received a similar answer. By this I mean, textile manufacturing improved leaps and bounds throughout the industrial revolution, however, despite their productivity, textile mills were some of the most dangerous places to work. Before I further explain my answer, let’s agree on a definition for data science. Wikipedia defines data science as, “an interdisciplinary field about scientific methods, processes, and systems to extract knowledge or insights from data in various forms, either structured or unstructured.” I see this as the process of acquiring, managing, and analyzing data. Advances in data science First, let's discuss why data science is definitely getting easier. Advances in technology and data collection have made data science easier. For one thing, data science as we know it wasn’t even possible 40 years ago, but due to advanced technology we can now analyze, gather, and manage data in completely new ways. Scripting languages like R and Python have mostly replaced more convoluted languages like Haskell and Fortran in the realm of data analysis. Tools like Hadoop bring together a lot of different functionality to expedite every element of data science. Smartphones and wearable tech collect data more effectively and efficiently than older data collection methods, which gives data scientists more data of higher quality to work with. Perhaps most importantly, the utility of data science has become more and more recognized throughout the broader world. This helps provide data scientists the support they need to be truly effective. These are just some of the reasons why data science is getting easier. Unintended consequences While many of these tools make data science easier in some respects, there are also some unintended consequences that actually might make data science harder. Improved data collection has been a boon for the data science industry, but using the data that is streaming in is similar to drinking out of a firehose. Data scientists are continually required to come up with more complicated ways of taking data in, because the stream of data has become incredibly strong. While R and Python are definitely easier to learn than older alternatives, neither language is usually accused of being parsimonious. What a skilled Haskell programming might be able to do in 100 lines, might take a less skilled Python scripter 500 lines. Hadoop, and tools like it, simplify the data science process, but it seems like there are 50 new tools like Hadoop a day. While these tools are powerful and useful, sometimes data scientists spend more time learning about tools and less time doing data science, just to keep up with the industry’s landscape. So, like many other fields related to computer science and programming, new tech is simultaneously making things easier and harder. Golden age of data science Let me rephrase the title question in an effort to provide even more illumination: is now the best time to be a data scientist or to become one? The answer to this question is a resounding yes. While all of the current drawbacks I brought up remain true, I believe that we are in a golden age of data science, for all of the reasons already mentioned, and more. We have more data than ever before and our data collection abilities are improving at an exponential rate. The current situation has gone so far as to create the necessity for a whole new field of data analysis, Big Data. Data science is one of the most vast and quickly expanding human frontiers at present. Part of the reason for this is what data science can be used for. Data science can effectively answer questions that were previously unanswered. Of course this makes for an attractive field of study from a research standpoint. One final note on whether or not data science is getting easier. If you are a person who actually creates new methods or techniques in data science, especially if you need to support these methods and techniques with formal mathematical and scientific reasoning, data science is definitely not getting easier for you. As I just mentioned, Big Data is a whole new field of data science created to deal with new problems caused by the efficacy of new data collection techniques. If you are a researcher or academic, all of this means a lot of work. Bootstrapped standard errors were used in data analysis before a formal proof of their legitimacy was created. Data science techniques might move at the speed of light, but formalizing and proving these techniques can literally take lifetimes. So if you are a researcher or academic, things will only get harder. If you are more of a practical data scientist, it may be slightly easier for now, but there’s always something! About the Author Erik Kappelman wears many hats including blogger, developer, data consultant, economist, and transportation planner. He lives in Helena, Montana and works for theDepartment of Transportation as a transportation demand modeler.

Tools to perform Deep Learning tasks are in abundance. You have programming languages that are adapted for the job or those specifically created to get the job done. Then, you have several frameworks and libraries which allow data scientists to design systems that sift through tonnes of data and learn from it. But a major challenge for all tools lies in tackling two primary issues: The size of the data The speed of computation Now, with petabytes and exabytes of data, it’s become way more taxing for researchers to handle. Take image processing for example. ImageNet itself is such a massive dataset consisting of trillions of images from several distinct classes that tackling this scale is a serious lip-biting affair. The speed at which researchers are able to get actionable insights from the data is also an important factor. Powerful hardware like multi-core GPUs, rumbling with raw power and begging to be tamed, have waltzed into the mosh pit of big data. You may try to humble these mean machines with old school machine learning stacks like R, SciPy or NumPy, but in vain. So, the deep learning community developed several powerful libraries to solve this problem, and they succeeded to an extent. But two major problems still existed - the frameworks failed to solve the problems of efficiency and flexibility together. This is where a one-of-a-kind, powerful, and flexible library like MXNet rises up to the challenge and makes developers’ lives a lot easier. What is MXNet? MXNet sits happy at over 10k stars on Github and has recently been inducted into the Apache Software Foundation. It focuses on accelerating the development and deployment of Deep Neural Networks at scale. This means exploiting the full potential of multi-core GPUs to process tonnes of data at blazing fast speeds. We’ll take a look at some of MXNet’s most interesting features over the next few minutes. Why is MXNET so good? Efficient MXNet is backed by a C++ backend which allows it to be extremely fast on even a single machine. It allows for automatically parallelizing computation across devices as well as synchronising the computation when multithreading is introduced. Moreover, the support for linear scaling means that not only the number of GPUs can be scaled, but MXNet also supports heavily distributed computing by scaling the number of machines as well. Moreover, MXNet has a graph optimisation layer that sits on top of a dynamic dependency scheduler, which enhances memory efficiency and speed. Extremely portable MXNet is extremely portable, especially as it can be programmed in umpteen languages such as C++, R, Python, Julia, JavaScript, Go, and more. It’s widely supported across most operating systems like Linux, Windows, iOS as well as Android making it multi-platform, including low level platforms. Moreover, it works well in cloud environments, like AWS - one of the reasons AWS has officially adopted MXNet as its deep learning framework of choice. You can now run MXNet from the Deep Learning AMI. Great for data visualization MXNet uses not only the mx.viz.plot_network method for visualising neural networks but it also has in-built support for Graphviz, to visualise neural nets as a computation graph. Check Joseph Paul Cohen’s blog for a great side-by-side visualisation of CNN architectures in MXNet. Alternatively, you could strip the TensorBoard off TensorFlow and use it with MXNet. Jump here for more info on that. Flexible MXNet supports both imperative and declarative/symbolic styles of programming, allowing you to blend both styles for increased efficiency. Libraries like Numpy and Torch support plain imperative programming, while TensorFlow, Theano, and Caffe support plain declarative programming. You can get a closer look at what these styles actually are here. MXNet is the only framework so far that mixes both styles to maximise efficiency and productivity. In-built profiler MXNet comes packaged with an in-built profiler that lets you profile execution times, layer-by-layer, in the network. Now, while you’ll be interested in using your own general profiling tools like gprof and nvprof to profile at kernel, function, or instruction level, the in-built profiler is specifically tuned to provide detailed information at a symbol or operator level. Limitations While MXNet has a host of attractive features that explain why it has earned public admiration, it has its share of limitations just like any other popular tool. One of the biggest issues encountered with MXNet is that it tends to give varied results when compile settings are modified. For example, a model would work well with cuDNN3 but wouldn’t with cuDNN4. To overcome issues like this, you might have to spend some time on forums. Moreover, it is a daunting task to write your own operators or layers in C++, to achieve efficiency. Although, with v0.9, the official documentation mentions that it has become easier. Finally, the documentation is introductory and is not organised well enough to create custom operators or to perform other advanced tasks. So, should I use MXNet? MXNet is the new kid on the block that supports modern deep learning models like CNNs and LSTMs. It boasts of immense speed, scalability, and flexibility to solve your deep learning problems and consumes as little as 4 Gigs of memory when running deep networks with almost a thousand layers. The core library including its dependencies can mash into a single C++ source file, which can be compiled on both Android and iOS, as well as a browser with the JavaScript extensions. But like all other libraries, it has it’s own hurdles, which aren’t that life threatening enough, to prevent one from getting the job done, and a good one at that! Is that enough to get you excited and start using MXNet? Go get working then! And don’t forget to tell us your experiences of working with MXNet.

Machine Learning is a powerful tool with applications in a wide variety of areas including image and object recognition, healthcare, language translation, and more. However, running ML tools requires complicated backends, complex architecture pipelines, and strict communication protocols. To overcome these obstacles, TensorFire, an in-browser DL library, is bringing the capabilities of machine learning to web browsers by running neural nets at blazingly fast speeds using GPU acceleration. It’s one more step towards democratizing machine learning using hardware and software already available with most people. How did in-browser deep learning libraries come to be? Deep Learning neural networks, a type of advanced machine learning, are probably one of the best approaches for predictive tasks. They are modular, can be tested efficiently and can be trained online. However, since neural nets make use of supervised learning (i.e. learning fixed mappings from input to output) they are useful only when large quantities of labelled training data and sufficient computational budget are available. They require installation of a variety of software, packages and libraries. Also, running a neural net has a suboptimal user experience as it opens a console window to show the execution of the net. This called for an environment that could make these models more accessible, transparent, and easy to customize. Browsers were a perfect choice as they are powerful, efficient, and have interactive UI frameworks. Deep Learning in-browser neural nets can be coded using JavaScript without any complex backend requirements. Once browsers came into play, in-browser deep learning libraries (read ConvNetJS, CaffeJS, MXNetJS etc.) have been growing in popularity. Many of these libraries work well. However, they leave a lot to be desired in terms of speed and easy access. TensorFire is the latest contestant in this race aiming to solve the problem of latency. What is TensorFire? It is a Javascript library which allows executing neural networks in web browsers without any setup or installation. It’s different from other existing in-browser libraries as it leverages the power of inbuilt GPUs of most modern devices to perform exhaustive calculations at much faster rates - almost 100x faster. Like TensorFlow, TensorFire is used to swiftly run ML & DL models. However, unlike TensorFlow which deploys ML models to one or more CPUs in a desktop, server, or mobile device, TensorFire utilizes GPUs irrespective of whether they support CUDA eliminating the need of any GPU-specific middleware. At its core, TensorFire is a JavaScript runtime and a DSL built on top of WebGL shader language for accelerating neural networks. Since, it runs in browsers, which are now used by almost everyone, it brings machine and deep learning capabilities to the masses. Why should you choose TensorFire? TensorFire is highly advantageous for running machine learning capabilities in the browsers due to four main reasons: 1.Speed They also utilize powerful GPUs (both AMD and Nvidia GPUs) built in modern devices to speed up the execution of neural networks. The WebGL shader language is used to easily write fast vectorized routines that operate on four-dimensional tensors. Unlike pure Javascript based libraries such as ConvNetJS, TensorFire uses WebGL shaders to run in parallel the computations needed to generate predictions from TensorFlow models. 2. Ease of use TensorFire also avoids shuffling of data between GPUs and CPUs by keeping as much data as possible on the GPU at a time, making it faster and easier to deploy.This means that even browsers that don’t fully support WebGL API extensions (such as the floating-point pixel types for textures) can be utilized to run deep neural networks.Since it has a low-precision approach, smaller models are easily deployed to the client resulting in fast prediction capabilities. TensorFire makes use of low-precision quantized tensors. 3. Privacy This is done by the website training a network on the server end and then distributing the weights to the client.This is a great fit for applications where the data is on the client-side and the deployment model is small.Instead of bringing data to the model, the model is delivered to users directly thus maintaining their privacy.TensorFire significantly improves latencies and simplifies the code bases on the server side since most computations happen on the client side. 4. Portability TensorFire eliminates the need for downloading, installing, and compiling anything as a trained model can be directly deployed into a web browser. It can also serve predictions locally from the browser. TensorFire eliminates the need to install native apps or make use of expensive compute farms. This means TensorFire based apps can have better reach among users. Is TensorFire really that good? TensorFire has its limitations. Using in-built browser GPUs for accelerating speed is both its boon and bane. Since GPUs are also responsible for handling the GUI of the computer, intensive GPU usage may render the browser unresponsive. Another issue is that although using TensorFire speeds up execution, it does not improve the compiling time. Also, the TensorFire library is restricted to inference building and as such cannot train models. However, it allows importing models pre-trained with Keras or TensorFlow. TensorFire is suitable for applications where the data is on the client-side and the deployed model is small. You can also use it in situations where the user doesn’t want to supply data to the servers. However, when both the trained model and the data are already established on the cloud, TensorFire has no additional benefit to offer. How is TensorFire being used in the real-world? TensorFire’s low-level APIs can be used for general purpose numerical computation running algorithms like PageRank for calculating relevance or Gaussian Elimination for inverting mathematical matrices in a fast and efficient way. Having capabilities of fast neural networks in the browsers allows for easy implementation of image recognition. TensorFire can be used to perform real-time client-side image recognition. It can also be used to run neural networks that apply the look and feel of one image into another, while making sure that the details of the original image are preserved. Deep Photo Style Transfer is an example. When compared with TensorFlow which required minutes to do the task, TensorFire took only few seconds. TensorFire also paves way for making tools and applications that can quickly parse and summarize long articles and perform sentiment analysis on their text. It can also enable running RNN in browsers to generate text with a character-by-character recurrent model. With TensorFire, neural nets running in browsers can be used for gesture recognition, distinguishing images, detecting objects etc. These techniques are generally employed using the SqueezeNet architecture - a small convolutional neural net that is highly accurate in its predictions with considerably fewer parameters. Neural networks in browsers can also be used for web-based games, or for user-modelling. This involves modelling some aspects of user behavior, or content of sites visited to provide a customized user experience. As TensorFire is written in JavaScript, it is readily available for use on the server side (available on Node.js) and thus can be used for server based applications as well. Since TensorFire is relatively new, its applications are just beginning to catch fire. With a plethora of features and advantages under its belt, TensorFire is poised to become the default choice for running in-browser neural networks. Because TensorFlow natively supports only CUDA, TensorFire may even outperform TensorFlow on computers that have non-Nvidia GPUs.

Cyber threats continue to cost companies money and reputation. Yet security seems to be undervalued. Or maybe it's just misunderstood. With a series of large-scale cyberattack events and the menace of ransomware, earlier WannaCry and now Petya, continuing to affect millions globally, it’s time you reimagined how your organization stays ahead of the game when it comes to software security.  Fortunately, machine learning can help support a more robust, reliable and efficient security initiative. Here are just 4 ways machine learning can support your software security strategy. Revamp of your company’s endpoint protection with machine learning We have seen in the past how a single gap in endpoint protection resulted in serious data breaches. In May this year, Mexican fast food giant Chipotle learned the hard way when cybercriminals exploited the company's point of sale systems to steal credit card information. The Chiptole incident was a very real reminder for many retailers to patch critical endpoints on a regular basis. It is crucial to guard your company’s endpoints which are virtual front doors to your organization’s precious information. Your cybersecurity strategy must consider a holistic endpoint protection strategy to secure against a variety of threats, both known and unknown. Traditional endpoint security approaches are proving to be ineffective and costing businesses millions in terms of poor detection and wasted time. The changing landscape of the cybersecurity market brings with it its own set of unique challenges (Palo Alto Networks have highlighted some of these challenges in their whitepaper here). Sophisticated Machine Learning techniques can help fight back threats that aren’t easy to defend with traditional ways. One could achieve this by adopting any of the three ML approaches: Supervised machine learning, unsupervised machine learning and reinforcement learning. Establishing the right machine learning approach entails a significant understanding of your expectations from the endpoint protection product. You might consider checking on the speed, accuracy, and efficiency of the machine learning based endpoint protection solution with the vendor to make an informed choice of what you are opting for. We recommend the use of a supervised machine learning approach for endpoint protection as it’s a proven way of malware detection and it delivers accurate results. The only catch is that these algorithms require relevant data in sufficient quantity to work on and the training rounds need to be speedy and effective to guarantee efficient malware detection. Some of the popular ML-based endpoint protection options available in the market are Symantec Endpoint Protection 14, CrowdStrike, and TrendMicro’s XGen. Use machine learning techniques to predict security threats based on historical data Predictive analytics is no longer just restricted to data science. By adopting predictive analytics, you can take a proactive approach to cybersecurity too. Predictive analytics makes it possible to not only identify infections and threats after they have caused damage, but also to raise an alarm for any future incidents or attacks. Predictive analytics is a crucial part of the learning process for the system. With sophisticated detection techniques the system can monitor network activities and report real-time data. One incredibly effective technique organizations are now beginning to use is a combination of  advanced predictive analytics with a red team approach. This enables organizations to think like the enemy and model a broad range of threats. This process mines and captures large sets of data which is then processed. The real value here is the ability to generate meaningful insights out of the large data set collected and then letting the red team work on processing and identifying potential threats. This is then used by the organization to evaluate its capabilities, to prepare for future threats and to mitigate potential risks. Harness the power of behavior analytics to detect security intrusions Behavior analytics is a highly trending area today in the cybersecurity space. Traditional systems such as antiviruses are skilled in identifying attacks based on historical data and matching signatures. Behavior analytics, on the other hand, detects anomalies and makes a judgement against what would be considered normal behaviour. As such, behavior analytics in enterprises is proving very effective when it comes to detecting intrusions that otherwise evade firewalls or antivirus software. It complements existing security measures such as firewall and antivirus rather than replacing them. Behavior analytics work well within private cloud and infrastructures and is able to detect threats within internal networks. One popular example is Enterprise Immune System, by the vendor Darktrace, which uses machine learning to detect abnormal behavior in the system. It helps IT staff narrow down their perimeter of search and look out for specific security events through a visual console. What’s really promising is that because Darktrace uses machine learning, the system is not just learning from events within internal systems, but from events happening globally as well. Use machine learning to close down IoT vulnerabilities Trying to manage large amounts of data and logs generated from millions of IoT devices manually could be overwhelming if your company relies on the Internet of Things. Many a time, IoT devices are directly connected to the network which means it is fairly easy for attackers and hackers to take advantage of your inadequately protected networks. It could therefore be next to impossible to build a secure IoT system, if you set out to identify and fix vulnerabilities manually. Machine learning can help you analyze and make sense of millions of data logs generated from IoT capable devices. Machine learning powered cybersecurity systems placed and seated directly inside your system can learn about security events as they happen. It can then monitor both incoming and outgoing IoT traffic in devices connected to the network and generate profiles for appropriate and inappropriate behavior inside your IoT ecosystem. This way the security system is able to react to even the slightest of irregularities and detect anomalies that were not experienced before. Currently, only a handful number of software and tools use Machine Learning or Artificial Intelligence for IoT security. But we are already seeing development on this front by major security vendors such as Symantec. Surveys carried out frequently on IoT continue to highlight security as a major barrier to IoT adoption and we are hopeful that Machine Learning will come to the rescue. Cyber crimes are evolving at a breakneck speed while businesses remain slow in adapting their IT security strategies to keep up with the times. Machine learning can help businesses make that leap to proactively address cyber threats and attacks by: Having an intelligent revamp of your company’s endpoint protection Investing in machine learning techniques that predict threats based on historical data Harnessing the power of behavior analytics to detect intrusions Using machine learning to close down IoT vulnerabilities And, that’s just the beginning. Have you used machine learning in your organization to enhance cybersecurity? Share with us your best practices and tips for using machine learning in cybersecurity in the comments below!

IoT contains hoard of sensors, vehicles and all devices that have embedded electronics which can communicate over the Internet. These IoT enabled devices generate tons of data every second. And with IoT Edge Analytics, these devices are getting much smarter - they can start or stop a request without any human intervention. 25 billion connected ”things” will be connected to the internet by 2020. - Gartner Research With so much data being generated by these devices, the question on everyone’s mind is: Will all this data be reliable and secure? When Brains meet Brawn: Blockchain for IoT Blockchain, an open distributed ledger, is highly secure and difficult to manipulate/corrupt by anyone connected over the network. It was initially designed for cryptocurrency based financial transactions. Bitcoin is a famous example which has Blockchain as its underlying technology. Blockchain has come a long way since then and can now be used to store anything of value. So why not save data in it? And this data will be secure just like every digital asset in a Blockchain is. Blockchain, decentralized and secured, is an ideal structure suited to form the underlying foundation for IoT data solutions. Current IoT devices and their data rely on the client service architecture. All devices are identified, authenticated, and connected via the cloud servers, which are capable of storing ample amount of data. But this requires huge infrastructure, which is all the more expensive. Blockchain not only provides an economical alternative but also since it works in a decentralized fashion it eliminates all single point of failures, creating a much secure and tougher network for IoT devices. This makes IoT more secure and reliable. Customers can therefore relax knowing their information is in safe hands. Today, Blockchain’s capabilities extend beyond processing financial transactions - It can now track billions of connected devices, process transactions and even co-ordinate between devices - a good fit for the IoT industry. Why Blockchain is perfect for IoT Inherent weak security features make IoT devices suspect. On the other hand, Blockchain with its tamper-proof ledger makes it hard to manipulate for malicious activities - thus, making it the right infrastructure for IoT solutions. Enhancing security through decentralization Blockchain makes it hard for intruders to intervene as it spans across a network of secure blocks. Change at a single location, therefore, does not affect the other blocks. The data or any value remains encrypted and is only visible to the person who has encrypted it using a private key. The cryptographic algorithms used in Blockchain technology ensure the IoT data remain private either for an individual organization or for the organizations connected in a network . Simplicity through autonomous 3rd-party-free transactions Blockchain technology is already a star in the finance sector thanks to the adoption of smart contracts, Bitcoin and other cryptocurrencies. Apart from providing a secured medium for financial transactions, it eliminates the need for third-party brokers such as banks to provide guarantee over peer-to-peer payment services. With Blockchain, IoT data can be treated in a similar manner, wherein smart contracts can be made between devices to exchange messages and data. This type of autonomy is possible because each node in the blockchain network can verify the validity of the transaction without relying on a centralized authority. Blockchain backed IoT solutions will thus enable trustworthy message sharing. Business partners can easily access and exchange confidential information within the IoT without a centralized management/regulatory authority. This means quicker transactions, lower costs and lesser opportunities for malicious intent such as data espionage. Blockchain's immutability for predicting IoT security vulnerabilities Blockchains maintain a history of all transactions made by smart devices connected within a particular network. This is possible because once you enter data in a Blockchain, it lives there forever in its immutable ledger. The possibilities for IoT solutions that leverage Blockchain’s immutability are limitless. Some obvious uses cases are more robust credit-scores and preventive health-care solutions that use data accumulated through wearables. For all the above reasons, we see significant Blockchain adoption by IoT based businesses in the near future.

When it comes to the ‘lingua franca’ of data science, there seems to be a face-off between R and Python. R has long been established as the language of researchers and statisticians but Python has come up quickly as a bona-fide challenger, helping embed analytics as a necessity for businesses and other organizations in 2017. If  a tech war does exist between the two languages, it’s a battle fought not so much on technical features but instead on the wider changes within modern business and technology. R is a language purpose-built for statistics, for performing accurate and intensive analysis. So, the fact that R is being challenged by Python — a language that is flexible, fast, and relatively easy to learn — suggests we are seeing a change in who’s actually doing data science, where they’re doing it, and what they’re trying to achieve. Python versus R — A Closer Look Let’s make a quick comparison of the two languages on aspects important to those working with data and see what we can learn about the two worlds where R and Python operate. Learning curve Python is the easier language to learn. While R certainly isn’t impenetrable, Python’s syntax marks it as a great language to learn even if you’re completely new to programming. The fact that such an easy language would come to rival R within data science indicates the pace at which the field is expanding. More and more people are taking on data-related roles, possibly without a great deal of programming knowledge — Python makes the barrier to entry much lower than R. That said, once you get to grips with the basics of R, it becomes relatively easier to learn the more advanced stuff. This is why statisticians and experienced programmers find R easier to use. Packages and libraries Many R packages are in-built. Python, meanwhile, depends upon a range of external packages. This obviously makes R much more efficient as a statistical tool — it means that if you’re using Python you need to know exactly what you’re trying to do and what external support you’re going to need. Data Visualization R is well-known for its excellent graphical capabilities. This makes it easy to present and communicate data in varied forms. For statisticians and researchers, the importance of that is obvious. It means you can perform your analysis and present your work in a way that is relatively seamless. The ggplot2 package in R, for example, allows you to create complex and elegant plots with ease and as a result, its popularity in the R community has increased over the years. Python also offers a wide range of libraries which can be used for effective data storytelling. The breadth of external packages available with Python means the scope of what’s possible is always expanding. Matplotlib has been a mainstay of Python data visualization. It’s also worth remarking on upcoming libraries like Seaborn. Seaborn is a neat little library that sits on top of Matplotlib, wrapping its functionality and giving you a neater API for specific applications. So, to sum up, you have sufficient options to perform your data visualization tasks effectively — using either R or Python! Analytics and Machine Learning Thanks to libraries like scikit-learn, Python helps you build machine learning systems with relative ease. This takes us back to the point about barrier to entry. If machine learning is upending how we use and understand data, it makes sense that more people want a piece of the action without having to put in too much effort. But Python also has another advantage; it’s great for creating web services where data can be uploaded by different people. In a world where accessibility and data empowerment have never been more important (i.e., where everyone takes an interest in data, not just the data team), this could prove crucial. With packages such as caret, MICE, and e1071, R too gives you the power to perform effective machine learning in order to get crucial insights out of your data. However, R falls short in comparison to Python, thanks to the latter’s superior libraries and more diverse use-cases. Deep Learning Both R and Python have libraries for deep learning. It’s much easier and more efficient with Python though — most likely because the Python world changes much more quickly, new libraries and tools springing up as quickly as the data science world hooks on to a new buzzword. Theano, and most recently Keras and TensorFlow have all made a huge impact on making it relatively easy to build incredibly complex and sophisticated deep learning systems. If you’re clued-up and experienced with R it shouldn’t be too hard to do the same, using libraries such as MXNetR, deepr, and H2O — that said, if you want to switch models, you may need to switch tools, which could be a bit of a headache. Big Data With Python, you can write efficient MapReduce applications with ease, or scale your R program on Hadoop to work with petabytes of data. Both R and Python are equally good when it comes to working with Big Data, as they can be seamlessly integrated with Big Data tools such as Apache Spark and Apache Hadoop, among many others. It’s likely that it’s in this field that we’re going to see R moving more and more into industry as businesses look for a concise way to handle large datasets. This is true in industries such as bioinformatics which have a close connection with the academic world and necessarily depend upon a combination of size and accuracy when it comes to working with data. So, where does this comparison leave us? Ultimately, what we see are two different languages offering great solutions to very different problems in data science. In Python, we have a flexible and adaptable language with a vibrant community of developers working on a huge range of problems and tasks, each one trying to find more effective and more intelligent ways of doing things. In R, we have a purely statistical language with a large repository of over 8000 packages for data analysis and visualization. While Python is production-ready and is better suited for organizations looking to harness technical innovation to its advantage, R’s analytical and data visualization capabilities can make your life as a statistician or data analyst easier. Recent surveys indicate that Python commands a higher salary than R — that is because it’s a language that can be used across domains; a problem-solving language. That’s not to say that R isn’t a valuable language; rather, Python is the language that just seems to fit the times at the moment. In the end, it all boils down to your background, and the kind of data problems you want to solve. If you come from a statistics or research background and your problems only revolve around statistical analysis and visualization, then R would best fit your bill. However, if you’re a Computer Science graduate looking to build a general-purpose, enterprise-wide data model which can integrate seamlessly with the other business workflows, you will find Python easier to use. R and Python are two different animals. Instead of comparing the two, maybe it’s time we understood where and how each can be best used and then harnessed their power to the fullest to solve our data problems. One thing is for sure, though — neither is going away anytime soon. Both R and Python occupy a large chunk of the data science market-share today, and it will take a major disruption to take either one of them out of the equation completely.

Do you have the motivation to learn machine learning? Given its relevance in today's landscape, you should be motivated to learn about this field. But if you're a web developer, how do you go about learning it? In this article, I show you how. So, let’s break this down. What is machine learning? You may be wondering why machine learning matters to you, or how you would even go about learning it. Machine learning is a smart way to create software that finds patterns in data without having to explicitly program for each condition. Sounds too good to be true? Well it is. Quite frankly many of the state-of-the-art solutions to the toughest machine learning problems don’t even come close to reaching 100 percent accuracy and precision. This might not sound right to you if you’ve been trained, or have learned, to be precise and deterministic with the solutions you provide to the web applications you’ve worked on. In fact, machine learning is such a challenging problem domain that data scientists describe problems to be tractable or not. Computer algorithms can solve tractable problems in a reasonable amount of time with a reasonable amount of resources, whereas, in-tractable problems simply can’t be solved. Decades more of R&D is needed at a deep theoretical level, to bring approaches and frameworks forward that will then take years to be applied and be useful to society. Did I scare you off? Nope? Okay great. Then you accept this challenge to learn machine learning.  But before we dive into how to learn machine learning, let's answer the question: Why does learning machine learning matter to you?  Well, you're a technologist and as a result, it’s your duty, your obligation, to be on the cutting edge. The technology world is moving at a fast clip and it’s accelerating. Take for example, the shortened duration between public accomplishments of machine learning versus top gaming experts. It took a while to get to the 2011 Watson v. Jeopardy champion, and far less time between AlphaGo and Libratus. So what's the significance to you and your professional software engineering career? Elementary dear my Watson—just like the so-called digital divide between non-technical and technical lay people, there is already the start of a technological divide between top systems engineers and the rest of the playing field in terms of making an impact and disrupting the way the world works.  Don’t believe me? When’s the last time you’ve programmed a self-driving car or a neural network that can guess your drawings? Making an impact and how to learn machine learning The toughest part about getting started with machine learning is figuring out what type of problem you have at hand because you run the risk of jumping to potential solutions too quickly before understanding the problem. Sure you can say this of any software design task, but this point can’t be stressed enough when thinking about how to get machines to recognize patterns in data. There are specific applications of machine learning algorithms that solve a very specific problem in a very specific way and it’s difficult to know how to solve a meta-problem if you haven’t studied the field from a conceptual standpoint. For me, a break through in learning machine learning came from taking Andrew Ng’s machine learning course on courser. So taking online courses can be a good way to start learning.  If you don’t have the time, you can learn about machine learning through numbers and images. Let's take a look.  Numbers Conceptually speaking, predicting a pattern in a single variable based on a direct—otherwise known as a linear relationship with another piece of data—is probably the easiest machine learning problem and solution to understand and implement.  The following script predicts the amount of data that will be created based on fitting a sample data set to a linear regression model: https://github.com/ctava/linearregression-go. Because there is somewhat of a fit of the sample data to a linear model, the machine learning program predicted that the data created in the fictitious Bob’s system will grow from 2017, 2018.  Bob’s Data 2017: 4401Bob’s Data 2018 Prediction: 5707  This is great news for Bob and for you. You see, machine learning isn’t so tough after all. I’d like to encourage you to save data for a single variable—also known as feature—to a CSV file and see if you can find that the data has a linear relationship with time. The following website is handy in calculating the number of dates between two dates: https://www.timeanddate.com/date/duration.html. Be sure to choose your starting day and year appropriately at the top of the file to fit your data. Images Machine learning on images is exciting! It’s fun to see what the computer comes up with in terms of pattern recognition, or image recognition. Here’s an example using computer vision to detect that grumpy cat is actually a Persian cat: https://github.com/ctava/tensorflow-go-imagerecognition. If setting up Tensorflow from source isn’t your thing, not to worry. Here’s a Docker image to start off with: https://github.com/ctava/tensorflow-go. Once you’ve followed the instructions in the readme.md file, simply:  Get github.com/ctava/tensorflow-go-imagerecognition Run main.go -dir=./ -image=./grumpycat.jpg Result: BEST MATCH: (66% likely) Persian cat Sure there is a whole discussion on this topic alone in terms of what Tensorflow is, what’s a tensor, and what’s image recognition. But I just wanted to spark your interest so that maybe you’ll start to look at the amazing advances in the computer vision field. Hopefully this has motivated you to learn more about machine learning based on reading about the recent advances in the field and seeing two simple examples of predicting numbers, and classifying images.I’d like to encourage you to keep up with data science in general. About the Author  Chris Tava is a Software Engineering / Product Leader with 20 years of experience delivering applications for B2C and B2B businesses. His specialties include: program strategy, product and project management, agile software engineering, resource management, recruiting, people development, business analysis, machine learning, ObjC / Swift, Golang, Python, Android, Java, and JavaScript.

Introduction and motivation In standard supervised machine learning, we need training data, i.e. a set of data points with known labels, and we build a model to learn the distinguishing properties that separate data points with different labels. This trained model can then be used to make label predictions for new data points. If we want to make predictions for another task (with different labels) in a different domain, we cannot use the model trained previously. We need to gather training data with the new task, and train a separate model. Transfer learning provides a framework to leverage the already existing model (based on some training data) in a related domain. We can transfer the knowledge gained in the previous model to the new domain (and data). For example, if we have built a model to detect pedestrians and vehicles in traffic images, and we wish to build a model for detecting pedestrians, cycles, and vehicles in the same data, we will have to train a new model with three classes because the previous model was trained to make two-class predictions. But we clearly have learned something in the two-class situation, e.g. discerning people walking from moving vechicles. In the transfer learning paradigm, we can use our learnings from the two-label classifier to the three-label classifier that we intend to construct. As such, we can already see that transfer learning has very high potential. In the words of Andrew Ng, a leading expert in machine learning, in his extremly popular NIPS 2016 tutorial, "Transfer learning will be next driver of machine learning success." Transfer learning in deep learning Transfer learning is particularly popular in deep learning. The reason for this is that it's very expensive to train deep neural networks, and they require huge amounts of data to be able to achieve their full potential. In fact, other recent successes of deep learning can be attributed to the availablity of a lot of data and stronger computational resources. But, other than a few large companies like Google, Facebook, IBM, and Microsoft, it's very difficult to accrue data and the computational machines required for training strong deep learning models. In such a situation, transfer learning comes to the rescue. Many pre-trained models, trained on a large amount of data, have been made available publically, along with the values of billions of parameters. You can use the pre-trained models on large data, and rely on transfer learning to build models for your specific case. Examples The most popular application of transfer learning is image classification using deep convolution neural networks (ConvNets). A bunch of high performing, state-of-the-art convolution neural network based image classifiers, trained on ImageNet data (1.2 million images with 100 categories), are available publically. Examples of such models include AlexNet, VGG16, VGG19, InceptionV3, and more, which takes months to train. I have personally used transfer learning to build image classifiers on top of VGG19 and InceptionV3. Another popular model is the pre-trained distributed word embeddings for millions of words, e.g word2vec, GloVe, FastText, etc. These are trained on all of Wikipedia, Google News, etc., and provide vector representations for a huge number of words. This can then be used in a text classification model. Strategies for transfer learning Transfer learning can be used in one the following four ways: Directly use pre-trained model: The pre-trained model can be directly used for a similar task. For example, you can use the InceptionV3 model by Google to make predictions about the categories of images. These models are already shown to have high accuracy. Fixed features: The knowledge gained in one model can be used to build features for the data points, and such features (fixed) are then fed to new models. For example, you can run the new images through a pre-trained ConvNet and the output of any layer can be used as a feature vector for this image. The features thus built can be used in a classifier for the desired situation. Similarly, you can directly use the word vectors in the text classification model. Fine-tuning the model: In this strategy, you can use the pre-trained network as your model while allowing for fine-tuning the network. For example, for the image classifier model, you can feed your images to the InceptionV3 model and use the pre-trained weights as an initialization (rather than random initialzation). The model will be trained on the much smaller user-provided data. The advantage of such a strategy is that weights can reach the global minima without much data and training. You can also make a portion (usually the begining layers) fixed, and only fine-tune the remaining layers. Combining models: Instead of re-training the top few layers of a pre-trained model, you can replace the top few layers by a new classifier, and train this combined network, while keeping the pre-trained portion fixed. Remarks It is not a good idea to fine-tune the pre-trained model if the data is too small and similar to the original data. This will result in overfitting. You can directly feed the data to the pre-trained model or train a simple classifier on the fixed features extracted from it. If the new data is large, it is a good idea to fine-tune the pre-trained model. In case the data is similar to the original, we can fine-tune only the top few layers, and fine-tuning will increase confidence in our predictions. If the data is very different, we will have to fine-tune the whole network. Conclusion Transfer learning allows someone without a large amount of data or computational capabilities to take advantage of the deep learning paradigm. It is an exciting research and application direction to use off-the-shelf pre-trained models and transfer them to novel domains.  About the Author  Janu Verma is a Researcher in the IBM T.J. Watson Research Center, New York. His research interests are in mathematics, machine learning, information visualization, computational biology and healthcare analytics. He has held research positions at Cornell University, Kansas State University, Tata Institute of Fundamental Research, Indian Institute of Science, and Indian Statistical Institute.  He has written papers for IEEE Vis, KDD, International Conference on HealthCare Informatics, Computer Graphics and Applications, Nature Genetics, IEEE Sensors Journals etc.  His current focus is on the development of visual analytics systems for prediction and understanding. He advises startups and companies on data science and machine learning in the Delhi-NCR area, email to schedule a meeting. Check out his personal website at http://jverma.github.io/.

What is keras? Keras is a high-level library for deep learning, which is built on top of Theano and Tensorflow. It is written in Python, and provides a scikit-learn type API for building neural networks. It enables developers to quickly build neural networks without worrying about the mathematical details of tensor algebra, optimization methods, and numerical techniques. The key idea behind Keras is to facilitate fast prototyping and experimentation. In the words of Francois Chollet, the creator of Keras: Being able to go from idea to result with the least possible delay is key to doing good research. This is a huge advantage, especially for scientists and beginner developers because one can jump right into deep learning without getting their hands dirty. Because deep learning is currently on the rise, the demand for people trained in deep learning is ever increasing. Organizations are trying to incorporate deep learning into their workflows, and Keras offers an easy API to test and build deep learning applications without considerable effort. Deep learning research is such a hot topic right now, scientists need a tool to quickly try out their ideas, and they would rather spend time on coming up with ideas than putting together a neural network model. I use Keras in my own research, and I know a lot of other researchers relying on Keras for its easy and flexible API. What are the key features of Keras? Keras is a high-level interface to Theano or Tensorflow and any of these can be used in the backend. It is extremely easy to switch from one backend to another. Training deep neural networks is a memory and time intensive task. Modern deep learning frameworks like Tensorflow, Caffe, Torch, etc. can also run on GPU, though there might be some overhead in setting up and running the GPU. Keras runs seamlessly on both CPU and GPU. Keras supports most of the neural layer types e.g. fully connected, convolution, pooling, recurrent, embedding, dropout, etc., which can be combined in any which way to build complex models. Keras is modular in nature in the sense that each component of a neural network model is a separate, standalone, fully-configurable module, and these modules can be combined to create new models. Essentially, layers, activation, optimizers, dropout, loss, etc. are all different modules that can be assembled to build models. A key advantage of modularity is that new features are easy to add as separate modules. This makes Keras fully expressive, extremely flexible, and well-suited for innovative research. Coding in Keras is extremely easy. The API is very user-friendly with the least amount of cognitive load. Keras is a full Python framework, and all coding is done in Python, which makes it easy to debug and explore. The coding style is very minimalistic, and operations are added in very intuitive python statements. How is Keras built? The core component of Keras architecture is a model. Essentially, a model is a neural network model with layers, activations, optimization, and loss. The simplest Keras model is Sequential, which is just a linear stack of layers; other layer arrangements can be formed using the Functional model. We first initialize a model, add layers to it one by one, each layer followed by its activation function (and regularization, if desired), and then the cost function is added to the model. The model is then compiled. A compiled model can be trained, using the simple API (model.fit()), and once trained the model can be used to make predictions (model.predict()). The similarity to scikit-learn API can be noted here. Two models can be combined sequentially or parallel. A model trained on some data can be saved as an HDF5 file, which can be loaded at a later time. This eliminates the need to train a model again and again; train once and make predictions whenever desired. Keras provides an API for most common types of layers. You can also merge or concatenate layers for a parallel model. It is also possible to write your own layers. Other ingredients of a neural network model like loss function, metric, optimization method, activation function, and regularization are all available with most common choices. Another very useful component of Keras is the preprocessing module with support for manipulating and processing image, text, and sequence data. A number of deep learning models and their weights obtained by training on a big dataset are made available. For example, we have VGG16, VGG19, InceptionV3, Xception, ResNet50 image recognition models with their weights after training on ImageNet data. These models can be used for direct prediction, feature building, and/or transfer learning. One of the greatest advantage of Keras is a huge list of example code available on the Keras GitHub repository (with discussions on accompanying blog) and on the wider Internet. You can learn how to use Keras for text classification using a LSTM model, generate inceptionistic art using deep dream, using pre-trained word embeddings, building variational autoencoder, or train a Siamese network, etc. There is also a visualization module, which provides functionality to draw a Keras model. This uses the graphviz library to plot and save the model graph to a file. All in all, Keras is a library worth exploring, if you haven't already.  Janu Verma is a Researcher in the IBM T.J. Watson Research Center, New York. His research interests are in mathematics, machine learning, information visualization, computational biology and healthcare analytics. He has held research positions at Cornell University, Kansas State University, Tata Institute of Fundamental Research, Indian Institute of Science, and Indian Statistical Institute.  He has written papers for IEEE Vis, KDD, International Conference on HealthCare Informatics, Computer Graphics and Applications, Nature Genetics, IEEE Sensors Journals etc.  His current focus is on the development of visual analytics systems for prediction and understanding. He advises startups and companies on data science and machine learning in the Delhi-NCR area, email to schedule a meeting. Check out his personal website here.

Like many cloud providers, IBM is invested and interested in developing and building out their machine learning offerings. Because more and more companies are interested in trying to figure out what machine learning can currently do and what it is on the cusp of doing, there has been a plethora of new services and companies trying to see if they can either do anything new or be better than the many other competitors in the market. While much of the machine learning world is based around cloud computing and the ability to horizontally scale during training, the reality is that not every company is cloud based. While IBM has had many machine learning tools and services available on their various cloud platforms, the IBM Machine Learning system seems to be just an on-premise compromise to many of the previously cloud-based APIs that IBM offered under their machine learning and Watson brand. While I am sure many enterprise-level companies have a large amount of data that would previously not have been able to use these services, I’m unaware whether and skeptical that these services would be any better or more utilitarian than their cloud counterparts. This seems like it may be particularly useful to companies in industries with many regulations or worries about hosting data outside of their private network, although this mentality seems like it is being slowly eroded away and becoming outdated in many senses. One of the great things about the IBM Machine Learning system (and many similar companies as well) are the APIs that allow developers to pick whatever language they would like and allowing multiple frameworks because there are so many available and interesting options at the moment. This is a really important aspect for something like deep learning, where there is a constant barrage of new architectures and ideas that iterate upon prior ideas, but require new architecture and developer implementations. While I have not used IBM’s new system and will most likely not be in any particular situation where it would be available, I was lucky enough to participate in IBM’s Catalyst program and used a lot of their cloud services for a deep learning project and testing out many of their available Bluemix offerings. While I thought the system was incredibly nice and simple to use compared to many other cloud providers, I found the various machine learning APIs they offered either weren’t that useful or seemed to provide worse results than their comparable Google, Azure, and other such services. Perhaps that has changed, or their new services are much better and will be available on this new system, but it is hard to find any definitive information that this is true. One aspect of these machine learning tools that I am not excited about is the constant focus on using machine learning to create some business cost-savings models (which the IBM Machine Learning press release touts), which companies may state is passed onto the customers, but it seems like this is rarely the truth (and one of the things that was stressed on in the IBM Machine Learning launch event). The ability for machine learning methodology to solve tedious and previously complicated problems is much more important and fascinating to me than simply saving money via optimizing business expenses. While many of these problems are much harder to solve and we are far from solving them, the current applications of machine learning in business and marketing areas often provides proof for the rhetoric that machine learning is exploitive and a toxic solution. Along with this, while the IBM Machine Learning and Data Science products may seem to be aimed at someone in a data scientist role, I can never help but wonder to what extent data scientists are actually using these tools outside of pure novelty or an extremely simple prototyping step in a more in-depth analytical pipeline. I personally think the ability of these tools to create usefulness for someone who may otherwise not be interested in many aspects of traditional data science is where the tools are incredibly powerful and useful. While not all traditional developers or analysts are interested in in-depth data exploration, creating pipelines, and many other traditional data science skills, the ability to do so at ease and without having to learn skills and tooling can be seen as a huge burden to those not particularly interested. While true machine learning and data science skills are unlikely to be completely replaceable, many of the skills that a traditional data scientist has will be less important as more people are capable of doing what previously may have been a quite technical or complicated process. These sorts of products and services that are a part of the IBM Machine Learning catalog reinforce that idea and are an important step in allowing services to be used regardless of data location or analytical background, and are an important step forward for machine learning and data science in general. About the author  Graham Annett is an NLP engineer at Kip (Kipthis.com). He has been interested in deep learning and has worked with and contributed to Keras. He can be found on GitHub or on his website. 

Ah, Raspberry Pi, the little computer that could. On its initial release back in 2012, it quickly gained popularity among creators and hobbyists as a cheap and portable computer that could be the brain of their hardware projects. Fast forward to 2017, and Raspberry Pi is on its third generation and has been used in many more projects across various fields of study.  Tech giants are noticing this trend and have started to pay closer attention to the miniature computer. Microsoft, for example, released Windows 10 IoT Core, a variant of Windows 10 that could run on a Raspberry Pi. Recently, Google revealed that they have plans to bring artificial intelligence tools to the Pi. And not just Google's AI, more and more AI libraries and tools are being ported to the Raspberry Pi every day.  But what does it all mean? Does it have any impact on Raspberry Pi’s usage? Does it change anything in the world of Internet of Things? For starters, let's recap what the Raspberry Pi is and how it has been used so far. The Raspberry Pi, in short, is a super cheap computer (it only costs $35) and is the size of a credit card. However, despite its ability to be usedasa usual, general-purpose computer, most people useRaspberry Pias the base of their hardware projects.  These projects range from simple toy-like projects to complicated gadgets that actually do important work. They can be as simple as a media center for your TV or as complex as a house automation system. Do keep in mind that this kind of projectscan always be built using desktop computers, but it's not really practical to do so without the low price and the small size of the Raspberry Pi.  Before we go on talking about having artificial intelligence on the Raspberry Pi, we need to have the same understanding of AI. (from http://i2.cdn.turner.com/cnn/2010/TECH/innovation/07/09/face.recognition.facebook/t1larg.tech.face.recognition.courtesy.jpg)  Artificial Intelligence has a wide range of complexity. It can range from a complicated digital assistant like Siri, to a news-sorting program, to a simple face detection system that can be found in many cameras. The more complicated the AI system, the bigger the computing power required by the system. So, with the limited processing power we have on the Raspberry Pi, the types of AI that can run on that mini computer will be limited to the simple ones as well.  Also, there's another aspect of AI called machine learning. It's the kind of technology that enables an AI to play and win against humans in a match of Go. The core of machine learning is basically to make a computer improve its own algorithm by processing a large amount of data. For example, if we feed a computer thousands of cat pictures, it will be able to define a pattern for 'cat' and use that pattern to find cats in other pictures.  There are two parts in machine learning. The first one is the training part, where we let a computer find an algorithm that suits the problem. The second aspect is the application part, where we apply the new algorithm to solve the actual problem. While the application part can usually be run on a Raspberry Pi, the training part requires a much higher processing power. To make it work, the training part is done on a high-performance computer elsewhere, and the Raspberry Pi only executes the training result. So, now we know that the Raspberry Pi can run simple AI. But what's the impact of this?  Well, to put it simply, having AI will enable creators to build an entirely new class of gadgets on the Raspberry Pi. It will allow the makers to create an actually smart device based on the small computer. Without AI, the so-called smart device will only act following a limited set of rules that have been defined. For example, we can develop a device that automatically turns off lights at a specific time every day, but without AI we can't have the device detect if there's anyone in the room or not.  With artificial intelligence, our devices will be able to adapt to unscripted changes in our environment. Imagine connecting a toy car with a Raspberry Pi and a webcam and have the car be able to smartly map its path to the goal, or a device that automatically opens the garage door if it sees our car coming in. Having AI on the Raspberry Pi will enable the development of such smart devices.  There's another thing to consider. One of Raspberry Pi's strong points is its versatility. With its USB ports and GPIO pins, the computer is able to interface with various digital sensors. The addition of AI will enable the Raspberry Pi to process even more sensors like fingerprint readers or speech recognition with a microphone, further enhancing its flexibility.  All in all, artificial intelligence is a perfect addition to the Raspberry Pi. It enables the creation of even smarter devices based on the computer and unlocks the potential of the Internet of Things to every maker and tinkerer in the world. About the author RakaMahesa is a game developer at Chocoarts (http://chocoarts.com/),who is interested in digital technology in general. Outside of work hours, he likes to work on his own projects, with Corridoom VR being his latest released game. Raka also regularly tweets as @legacy99. 

If you have pitched the idea of a set of APIs to your boss, you might have run across this question. “Why do we need an API, and what does it have to do with an economy?” The answer is the API economy - but it's more than likely that that is going to be met with more questions. So let's take some time to unpack the concept and get through some of the hyperbole surrounding the topic. An economy (From Greek οίκος – "household" and νέμoμαι – "manage") is an area of the production, distribution, or trade, and consumption of goods and services by different agents in a given geographical location. - Wikipedia If we take the definition of economy from Wikipedia and the definition of API as an Application Programming Interface, then what we should be striving to create is a platform (as the producer of the API) that will attract a set of agents that will use that platform to create, trade or distribute goods and services to other agents over the Internet (our geography has expanded). The central tenet of this economy is that the APIs themselves need to provide the right set of goods (data, transactions, and so on) to attract other agents (developers and business partners) that can grow their businesses alongside ours and further expand the economy. This piece from Gartner explains the API economy very well. This is a great way of summing it up: "The API economy is an enabler for turning a business or organization into a platform." Let’s explore a bit more about APIs and look at a few examples of companies that are doing a good job of running API platforms. The evolution of the API economy If you asked someone what an API actually was 10 or more years ago, you might have received puzzled looks. The Application Programming Interface at that time was something that the professional software developer was using to interface with more traditional enterprise software. That evolved into the popularity of the SDK (Software Development Kit) and a better mainstream understanding of what it meant to create integrations or applications on pre-existing platforms. Think of the iOS SDK or Android SDK and how those kits and the distribution channels that Apple and Google created have led to the explosion of the apps marketplace. Jeff Bezos’s mandate that all IT assets have an API at Amazon was a major event in the API economy timeline. Amazon continued to build APIs such as SNS, SQS, Dynamo and many others. Each of these API components provided a well-defined service that any application can use and has significantly reduced the barrier to entry for new software and service companies. With this foundation set, the list of companies providing deep API platforms has steadily increased. How exactly does one profit in the API economy? If we survey a small set of API frameworks, we can see that companies use their APIs in different ways to add value to their underlying set of goods or create a completely new revenue stream for the company. Amazon AWS Amazon AWS is the clearest example of an API as a product unto itself. Amazon makes available a large set of services that provide defined functionality and for which Amazon charges with rates based upon usage of CPU and storage (it gets complicated). Each new service they launch addresses a new area of need and work to provide integrations between the various services. Social APIs Facebook, Twitter and others in the social space, run API platforms to increase the usage of their properties. Some of the inherent value in Facebook comes from sites and applications far afield from facebook.com and their API platform enables this. Twitter has had a more complicated relationship with its API users over time, but the API does provide many methods that allow both apps and websites to tap into Twitter content and thus extend Twitter’s reach and audience size. Chat APIs Slack has created a large economy of applications focused around its chat services and built up a large number of partners and smaller applications that add value to the platform. Slack’s API approach is one that is centered on providing a platform for others to integrate with and add content into the Slack data system. This approach is more open than the one taken by Twitter and the fast adoption has added large sums to Slack’s current valuation. Along side the meteoric rise of Slack, the concept of the bot as an assistant has also taken off. Companies like api.ai are offering services to enable chat services with AI as a service. The service offerings that surround the bot space are growing rapidly and offer a good set of examples as to how a company can monetize their API. Stripe Stripe competes in the payments as a service space along with PayPal, Square and Braintree. Each of these companies offers API platforms that vastly simplify the integration of payments into web sites and applications. Anyone who has built an e-commerce site before 2000 can and will appreciate the simplicity and power that the API economy brings to the payment industry. The pricing strategy in this space is generally on a per use case and is relatively straightforward. It Takes a Community to make the API economy work There are very few companies that will succeed by building an API platform without growing an active community of developers and partners around it. While it is technically easy to create and API given the tooling available, without an active support mechanism and detailed and easily consumable documentation your developer community may never materialize. Facebook and AWS are great examples to follow here. They both actively engage with their developer communities and deliver rich sets of documentation and use-cases for their APIs.

MongoDB, founded in 2007 with more than 15 million downloads, excels at supporting real-time analytics for big data applications. Rather than storing data in tables made out of individual rows, MongoDB stores it in collections made out of JSON documents. But, why use MongoDB? How does it work? What issues should you pay attention to? Let’s answer these questions in this post. MongoDB, a NoSQL database MongoDB is a NoSQL database, and NoSQL == Not Only SQL. The data structure is combined with keyvalue, like JSON. The data type is very flexible, but flexibility can be a problem if it’s not defined properly. Here are some good reasons you should use MongoDB: If you are a front-end developer, MongoDB is much easier to learn than mySQL, because the MongoDB base language is JavaScript and JSON. MongoDB works well for big data, because for instance, you can de-normalize and flatten 6 tables into just 2 tables. MongoDB is document-based. So it is good to use if you have a lot of single types of documents. So, now let’s examine how MongoDB works, starting with installing MongoDB: Download MongoDB from https://www.mongodb.com/download-center#community. De-zip your MongoDB file. Create a folder for the database, for example, Data/mydb. Open cmd to the MongoDB path, $ mongod –dbpath ../data/mydb. $ mongo , to make sure that it works. $ show dbs, and you can see two databases: admin and local. If you need to shut down the server, use $db.shutdownServer(). MongoDB basic usage Now that you have MongoDB on your system, let’s examine some basic usage of MongoDB, covering insertion of a document, removal of a document, and how to drop a collection from MongoDB. To insert a document, use the cmd call. Here we use employee as an example to insert a name, an account, and a country. You will see the data shown in JSON: To remove the document: db.collection.remove({ condition }), justOne) justOne: true | false, set to remove the first data, but if you want to remove them all, use db.employee.remove({}). To drop a collection (containing multiple documents) from the database, use: db.collection.drop() For more commands, please look at the MongoDB documentation. What to avoid Let’s examine some points that you should note when using MongoDB: Not easy to change to another database: When you choose MongoDB, it isn’t like other RDBMSes. It can be difficult to change, for example, from MongoDB to Couchbase. No support for ACID: ACID (Atomicity, Consistency, Isolation, Durability) is the basic item of transactions, but most NoSQL databases don’t guarantee ACID, so you need more technical skills in order to do this. No support for JOIN: Since the NoSQL database is non-relational, it does not support JOIN. Document limited: MongoDB uses stock data in documents, and these documents are in JSON format. Because of this, MongoDB has a limited data size, and the latest version supports up to 16 MB per document. Filter search has to correctly define lowercase/uppercase: For example: db.people.find（{name: ‘Russell’}） and db.people.find（{name: ‘ russell’}） are different. You can filter by regex, such as db.people.find（{name:/Russell/i}）, but this will affect performance. I hope this post has provided you with some important points about MongoDB which will help you decide if you have a big data solution that is a good fit for using this NoSQL database. About the author  Tess Hsu is a UI design and front-end programmer. He can be found on GitHub.

“Deep learning (deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using multiple processing layers, with complex structures or otherwise, composed of multiple nonlinear transformations.” -Wikipedia High Performance Computing is not a new concept, but only recently the technical advances along with economies of scale have ensured that HPC is accessible to the masses with affordable yet powerful configurations. Anyone interested can buy commodity hardware and start working on Deep Learning, thus bringing a machine learning subset of artificial intelligence out of research labs and into garages. DeepLearning.net is a starting point for more information about Deep Learning. Nvidia ParallelForAll is a nice resource for learning GPU-based Deep Learning (Core Concepts, History and Training, and Sequence Learning). What is CNTK Microsoft Research released its Computational Network Toolkit in January this year. CNTK is a unified deep-learning toolkit that describes neural networks as a series of computational steps via a directed graph. CNTK allows the following models: -          Feed Forward Deep Neural Networks (DNN) -          Convolutional Neural Networks (CNN) -          Recurrent Neural Networks (RNN)/Long Short Term Memory Units (LTSM) -          Stochastic Gradient Descent (SGD) Why CNTK? Better Scaling When Microsoft CNTK was released, the stunning feature that it brought was distributed computing, that is, a developer was not limited by the number of GPUs installed on a single machine. This was a significant breakthrough because even the best of machines was limited by4-way SLI, thus limiting the total number of cores to 4 x 3072 = 12288. The configuration of the developer machine put an extra load on the hardware configuration, because this configuration left very little room for upgrades. There is only one motherboard available that supports4-way PCI-E Gen3x16, and there are very few manufacturers who provide good quality 1600W watt power supply to support four Titans. This meant that developers were forced to pay a hefty premium for upgradability in terms of the motherboard and processor, settling for an older generation processor. Distributed computing in High Performance Computing is essential, since it allows scaling out as opposed to scaling up. Developers can build grids with cheaper nodes and the latest processors with a lower hardware cost for an entry barrier. Microsoft Research demonstrated in December 2015 that distributed GPU computing is most efficient in CNTK. In comparison, Google TensorFlow, FAIR Torch, andCaffe did not allow scaling beyond a single machine, and Theano was the worst, as it did not even scale on multiple GPUs on the same machine. Google Research, on April 13, released support for distributed computing. The speed up claimed is 56X for 100 GPUs and 40X for 50 GPUs. The performance deceleration is sharp for any sizable distributed Machine Learning setup. I do not have any comparative performance figures for CNTK, but scaling with GPUs on a single machine for CNTKhad very good numbers. GPU Performance One of the shocking finds with my custom build commodity hardware (2xTitanX) was the TFLOPS achieved under Ubuntu 14.04 LTS and Windows 10. With the fully updated OS and latest drivers from NVIDIA, I got double the number of TFLOPS under Windows than Ubuntu. I would like to rerun the samples with Ubuntu 16.04 LTS, but until then, I have a clear winner in performance with Windows. CNTK works perfectly on Windows, but TensorFlow has a dependency onBazel, which as of now does not build on Windows (Bug#947). Google can look into this and make TensorFlow work on windows, or Ubuntu &Nvidia can achieve the same TFLOPS as Windows. But until that time, architects have two options:to either settle for lower TFLPOS under Ubuntu with TensorFlow, or migrate to CNTK with increased performance. Getting started with CNTK Let’s see how toget started with CNTK. Binary Installation Currently, the CNTK binary installation is the easiest way to get started with CNTK. Just follow the instructions. The only downside is that the currently available binaries are compiled with CUDA 7.0, rather than the latest CUDA 7.5 (released almost a year ago). Codebase Compilation If you want to learn CNTK in detail, and if you are feeling adventurous, you should try compiling CNTK from source. Compile the code base, even when you do not expect to use the generated binary, because the whole compilation process will be a good peek under the hood and enhance your understanding of Deep Learning. The instructions for the Windows installation are available here, whereas the Linux installation instructions are available here. If you want to enable 1-bit Stochastic Gradient Descent (1bit-SGD), you should follow these instructions.1bit-SGD is licensed more restrictively, and you have to understand the differences if you are looking for commercial deployments. Windows Compilation is characterized by older versions of libraries. NvidiaCUDA andcuDNN were recently updated to 7.5, whereas other dependencies such asNvidia CUB, Boost, and OpenCV are still using older versions. Kindly pay extra attention to the versions listed in the documentation to ensure smooth compilation. Nvidia has updated the support for its Nsight to Visual Studio 2015; however, Microsoft CNTK still supportsonly Visual Studio 2013. Samples To test the CNTK installation, here are some really great samples: Simple2d (Feed Forward) Speech / AN4 (Feed Forward &LSTM) Image / MNIST (CNN) Text / PennTreeback (RNN) Alternative Deep Learning Toolkits Theano Theano is possibly the oldest Deep Learning Framework available. The latest release, 0.8,which was released on March 16, enables the much awaited multi-GPU support (there are no indications of distributed computing support though).cuDNNv5 and CNMeM are also supported. A detailed report is available here. Python bindings are available. Caffe Caffe is a deep learning framework primarily oriented towards image processing. Python bindings are available. Google TensorFlow TensorFlow is a deep learning framework written in C++ with Python API bindings. The computation graph is pure Python, making it slower than other frameworks, as demonstrated by benchmarks. Google has been pushing Go for a long time now, and it has even open sourced the language. But when it came to TensorFlow, Python was chosen over Go. There are concerns about Google supporting commercial implementations. FAIR Torch Facebook AI Research (Fair) has release its extension to Torch7. Torch is a scientific computing framework with Lua as its primary language. Lua has certain advantages over Python (lower interpreter overhead, simpler integration with C code), which lend themselves to Torch. Moreover, multi-core using OpenMP directives points to better performance. Leaf Leaf is the latest addition to the machine learning framework. It is based on the Rust programming language (supposed to replace C/C++). Leaf is a framework created by hackers for hackers rather than scientists. Leaf has some nice performance improvements. Conclusion Deep Learning with GPUs is an emerging field, and there is much required to be done to make good products out of machine learning. So every product needs to evaluate all of the possible alternatives (programming language, operating system, drivers, libraries, and frameworks) available for specific use-cases. Currently, there is no one-size-fits-all approach available. About the author SarvexJatasra is a Technology Aficionado, exploring ways to apply technology to make lives easier. He is currently working as the Chief Technology Officer at 8Minutes. When not in touch with technology, he is involved in physical activities such as swimming and cycling.