This free 3-hour Training on AI & ChatGPT (worth $399) will help you become a master of 20+ AI tools & prompting techniques and save 16 hours/week.

Get it now for absolutely free! (for first 100 users only) 🎁

You will learn how to:

➣ Build business that make $10,000 by just using AI tools

➣ Make quick & smarter decisions using AI-led data insights

➣ Write emails, content & more in seconds using AI

➣ Solve complex problems, research 10x faster & save 16 hours every week

Sponsored

Welcome to DataPro #118 – Your Weekly Data Science & ML Wizardry! 🌟

Stay sharp in the fast-evolving world of data science with this week’s essential strategies, tools, and trends. We’ve handpicked the best to supercharge your projects, refine accuracy, and amp up performance. Ready for this week’s power-ups? Let’s go!

🚨 Packt Conference Alert! 🚨

Stay at the forefront of AI innovation! 🚀 Join us for 3 action-packed days of LIVE sessions with 20+ top experts and unleash the full power of Generative AI at our upcoming conference. Don’t miss out - Claim your spot today!

🔍 Algorithm Insight: Model of the Week Unveiled

➣Gemini Models Hit GitHub Copilot: Dive into code generation like never before with Gemini models, now integrated in GitHub Copilot through Google Cloud’s partnership.

➣SimpleQA from OpenAI: A new benchmark tool to measure the factual accuracy of language models.

➣Theory of Mind in AI: Evaluating the latest with SimpleToM, a new tool testing language models’ understanding of human perspectives.

➣Meta AI’s LongVU: Tackling long video comprehension with a new multimodal language model.

➣JetBrains Introduces CoqPilot: A Plugin for LLM-Based Proof Generation.

➣Jupyter Releaser: Streamlining software releases for Jupyter tools just got easier.

🚀 Tech Trend Radar: What's Making Waves?

➣LLMs for Chunked Retrieval: How to leverage LLMs for smarter, chunk-based information recall.

➣OmniParser by Microsoft AI: Convert UI screenshots to structured data on Hugging Face.

➣Hawkish 8B Financial Model: Outperforming in finance tests, this model aces CFA Level 1 exams.

➣Gen-AI Safety Stack: A guide to safety strategies for text-to-image model applications.

➣Equation Solving in Python: A must-read on closed-form versus numerical solutions.

🛠️ Tool Time: Comparing Platforms & Services

➣Cohere’s Aya Expanse: A powerful multilingual model suite closing the language gap in AI.

➣Meta AI’s NotebookLlama: An open-source alternative to Google’s NotebookLM, now available.

➣AI for Screen Interaction: Explore Claude 3.5’s new screen navigation capabilities.

➣Text Embeddings with Amazon RDS & Bedrock: Seamlessly embed and retrieve text data from Amazon RDS using Amazon’s Bedrock.

➣Custom Observability Solution: Track, log, and improve generative AI applications with Bedrock.

📊 Real-World Impact: Success Stories & Case Studies

➣Python One-Liners for Data Cleaning: 10 concise solutions for everyday data wrangling.

➣2024’s Top Python Libraries: Must-have Python tools for data science this year.

➣Automating Model Selection with LLMs: Streamlining model testing and tuning.

➣5 Tips to Optimize Language Models: Quick techniques for better model performance.

➣Lessons Beyond AI: Three crucial takeaways from a recent data science conference.

🌍 ML Newsflash: Industry Discoveries & Updates

➣Hugging Face Models on Mobile: A step-by-step guide to deploying Hugging Face models on mobile.

➣Python for Proximity Mapping: Learn how to create distance maps in Python for quick insights.

➣Data Leakage Alert: Key practices to prevent leaks during data preprocessing.

➣In-Depth RAG Guide: Understand Retrieval Augmented Generation with a breakdown of each component.

➣Beyond Basic Attention in Transformers: Analyzing positional embedding techniques for improved model accuracy.

Dive into this week’s DataPro and stay on top of everything that’s shaping the world of Data Science & Machine Learning!

Take our weekly survey and get a free PDF copy of our best-selling book,"Interactive Data Visualization with Python - Second Edition."We appreciate your input and hope you enjoy the book!

Share Your Insights and Shine! 🌟💬

Cheers,

Merlyn Shelley,

Editor-in-Chief, Packt.

➽ RAG-Driven Generative AI: This new title, RAG-Driven Generative AI, is perfect for engineers and database developers looking to build AI systems that give accurate, reliable answers by connecting responses to their source documents. It helps you reduce hallucinations, balance cost and performance, and improve accuracy using real-time feedback and tools like Pinecone and Deep Lake. By the end, you’ll know how to design AI that makes smart decisions based on real-world data—perfect for scaling projects and staying competitive! Start your free trial for access, renewing at $19.99/month.

➽ Building Production-Grade Web Applications with Supabase: This new book is all about helping you master Supabase and Next.js to build scalable, secure web apps. It’s perfect for solving tech challenges like real-time data handling, file storage, and enhancing app security. You'll even learn how to automate tasks and work with multi-tenant systems, making your projects more efficient. By the end, you'll be a Supabase pro! Start your free trial for access, renewing at $19.99/month.

➽ Python Data Cleaning and Preparation Best Practices: This new book is a great guide for improving data quality and handling. It helps solve common tech issues like messy, incomplete data and missing out on insights from unstructured data. You’ll learn how to clean, validate, and transform both structured and unstructured data—think text, images, and audio—making your data pipelines reliable and your results more meaningful. Perfect for sharpening your data skills! Start your free trial for access, renewing at $19.99/month.

➽ Gemini Models on GitHub Copilot: GitHub and Google Cloud’s partnership introduces Gemini 1.5 Pro to GitHub, enhancing AI-driven code generation, analysis, and optimization for developers. The Gemini model, with a two-million-token context window, will integrate into GitHub Copilot, Google AI Studio, Vertex AI, and popular IDEs.

➽ OpenAI Introduces SimpleQA: AI Benchmark for Measuring the Factuality of Language Models. The blog introduces SimpleQA, a factuality benchmark for evaluating how accurately language models answer short, fact-seeking questions. SimpleQA emphasizes correctness, topic diversity, and difficulty for advanced models. Built with rigorous quality checks, it helps researchers gauge model performance and reduce “hallucinations” in AI responses.

➽ SimpleToM: Evaluating Applied Theory of Mind Capabilities in Large Language Models. The blog discusses SimpleToM, a dataset developed to assess Theory of Mind (ToM) in large language models (LLMs) through realistic scenarios. Unlike prior methods, it evaluates nuanced mental state inferences and behavior judgments, revealing gaps in LLMs’ understanding and application of social reasoning in real-world situations.

➽ Data Minimization Does Not Guarantee Privacy: The blog explains the data minimization principle in machine learning, emphasizing the need to collect only essential data to reduce privacy risks, as outlined by global data protection laws. It discusses challenges in operationalizing this principle due to inherent data correlations and highlights privacy audits, using adversarial attacks, to identify vulnerabilities.

➽ Meta AI Releases LongVU: A Multimodal Large Language Model that can Address the Significant Challenge of Long Video Understanding. The blog highlights Meta AI's release of LongVU, a Multimodal Large Language Model designed to tackle the challenges of long video understanding. By using adaptive compression techniques and cross-modal queries, LongVU reduces redundant frames and tokens, enabling efficient processing of hour-long videos within limited context lengths, thereby advancing video analysis in AI.

➽ JetBrains Researchers Introduce CoqPilot: A Plugin for LLM-Based Generation of Proofs. The blog introduces CoqPilot, a VS Code extension from JetBrains that automates Coq proof generation. By using LLMs like GPT-4 and tools like CoqHammer, CoqPilot fills proof gaps, verifies solutions, and replaces incomplete proofs. This integration streamlines proof creation, enhancing efficiency in software reliability and formal verification tasks.

➽ Jupyter Releaser: Streamlining Software Releases for the Jupyter Ecosystem. The blog covers Jupyter Releaser, a tool launched by the Jupyter team to streamline release management across Jupyter projects. By automating tasks like changelog creation and artifact publishing via GitHub Actions, Jupyter Releaser reduces errors, speeds up releases, and promotes consistency, benefiting the broader open-source development community.

➽ How and Why to Use LLMs for Chunk-Based Information Retrieval. The article explores using Large Language Models (LLMs) like GPT-4 for chunk-based information retrieval. By utilizing hybrid search techniques—combining term frequency algorithms and vector-based search—LLMs identify relevant text chunks. Despite improving retrieval, issues like irrelevant chunk selection persist, potentially misleading LLM responses in systems like RAG (Retrieval-Augmented Generation).

➽ Microsoft AI Releases OmniParser Model on HuggingFace: A Compact Screen Parsing Module that can Convert UI Screenshots into Structured Elements. OmniParser by Microsoft enables GUI interaction for AI by interpreting interface elements from screenshots without HTML or metadata. Using vision-based detection, icon description, and OCR, it enhances AI usability across platforms, boosting accuracy in interface tasks and advancing applications in automation and accessibility.

➽ Meet Hawkish 8B: A New Financial Domain Model that can Pass CFA Level 1 and Outperform Meta Llama-3.1-8B-Instruct in Math & Finance Benchmarks. The article introduces Hawkish 8B, a finance-focused AI model excelling in financial analysis and quantitative tasks. With specialized training in economics and market analysis, Hawkish 8B surpasses other models in benchmarks and even passes CFA Level 1, aiding finance professionals.

➽ Gen-AI Safety Landscape: A Guide to the Mitigation Stack for Text-to-Image Models: The article covers Text-to-Image (T2I) AI models like Latent Diffusion Models, detailing capabilities like inpainting and associated risks, including generating inappropriate content. It emphasizes a robust safety mitigation stack across training, fine-tuning, and post-deployment to minimize harmful outputs and ethical concerns.

➽ Solving Equations in Python: Closed-Form vs Numerical: The article explores when closed-form solutions are possible in mathematical models, such as Kepler’s orbital equation, and why numerical methods are often needed. Using Python’s SymPy, it examines equations to build intuition around solvable forms and complexities that defy simple algebraic solutions.

➽ Demystifying Azure Storage Account Network Access: The article details network access control for Azure storage accounts within medallion architecture, focusing on using service endpoints and private endpoints. It explains setup configurations, firewall rules, and network security groups (NSGs) to securely enable data access for virtual machines while preventing unauthorized access.

➽ Cohere for AI Releases Aya Expanse (8B & 32B): A State-of-the-Art Multilingual Family of Models to Bridge the Language Gap in AI. The article introduces Aya Expanse by Cohere for AI, an open-weight, multilingual language model family addressing underrepresentation in NLP. Designed to support low-resource languages, Aya Expanse achieves high accuracy on multilingual benchmarks, promoting inclusivity and equitable access to AI-driven tools across diverse linguistic communities.

➽ Meta AI Silently Releases NotebookLlama: An Open Version of Google's NotebookLM. The article introduces Meta's NotebookLlama, an open-source alternative to Google’s NotebookLM, integrating LLMs into a notebook interface for accessible, scalable data analysis and documentation. NotebookLlama offers customizable deployment, enhances code-writing and documentation, and empowers the AI community with a flexible, community-driven tool.

➽ Computer Use and AI Agents: A New Paradigm for Screen Interaction: The article explores recent advancements in multimodal AI agents from Anthropic, Microsoft, and Apple. These agents enhance computer and mobile screen interaction using technologies like Anthropic’s Claude 3.5, Microsoft’s OmniParser, and Apple’s Ferret-UI, highlighting varied approaches for parsing screens and performing actions, albeit with ongoing challenges.

➽ Embed textual data in Amazon RDS for SQL Server using Amazon Bedrock: The article explains how to generate vector embeddings from Wikipedia data stored in an Amazon RDS SQL Server database. Using Amazon Bedrock and Amazon SageMaker, the solution integrates embeddings into SQL Server for similarity search in generative AI applications, streamlining analysis through AWS’s managed AI services.

➽ Empower your generative AI application with a comprehensive custom observability solution: The article introduces an observability and evaluation solution for Amazon Bedrock to enhance generative AI applications. By integrating decorators in application code, this solution captures logs and metrics, supporting Retrieval Augmented Generation (RAG) evaluations and enabling proactive monitoring, quality improvement, and secure data handling across AI workflows.

➽ 10 Useful Python One-Liners for Data Cleaning: The article provides Python one-liners for common data cleaning tasks like handling duplicates, validating formats, managing missing values, and scaling numbers. It guides users in cleaning a sample dataset to prepare it for analysis, covering essentials like email validation, date standardization, and whitespace trimming.

➽ 10 Essential Python Libraries for Data Science in 2024: The article covers ten essential Python libraries for data science, each specializing in a critical task like data collection (Scrapy), manipulation (pandas), visualization (Matplotlib), machine learning (scikit-learn), and deployment (Flask). These libraries streamline end-to-end workflows, making data science more accessible and efficient.

➽ Selection and Experimentation Automation with LLMs: The article demonstrates how to automate model selection and experimentation using large language models (LLMs). By applying LLMs like GPT-4 with Scikit-Learn, the code automates model evaluation, selects the best-performing model, and even suggests hyperparameters for tuning. This approach streamlines model experimentation in data science.

➽ 5 Tips for Optimizing Language Models: The article provides five essential tips for optimizing language models: using prompt engineering to refine model responses, applying Retrieval Augmented Generation (RAG) for contextual accuracy, fine-tuning for task specificity, adjusting hyperparameters to enhance performance, and compressing models for efficiency and accessibility across various platforms.

➽ Three Crucial Data Lessons That I Learned from a Data Conference That’s Not Related to AI. The article shares insights from a data conference, emphasizing cost control, effective data translation, and cross-department collaboration to boost data team ROI. Practical tips include using cost-monitoring dashboards, fostering data literacy, and aligning data projects with strategic business goals.

➽ How Prefab scales with Spanner’s PostrgeSQL interface: Prefab uses Google Cloud Spanner’s PostgreSQL interface for its impressive scalability, simplicity, and cost-effectiveness. Spanner offers the robustness of PostgreSQL with high availability, strong ACID compliance, and horizontal scaling, making it ideal for Prefab's feature flagging and dynamic logging services.

➽ How to Deploy Hugging Face Models on Mobile Devices: This guide covers deploying Hugging Face models on mobile by converting models like DistilBERT into ONNX format, then quantizing to reduce file size for mobile compatibility. The article also demonstrates testing and setup for Android deployment, enabling efficient and scalable use of machine learning on mobile devices.

➽ Building Interactive Data Science Applications with Python:This article details building interactive data science applications using Python libraries like Streamlit, Gradio, Dash, and Panel. It explains creating engaging apps with features like user inputs, feedback, and multimedia elements, and includes an example dashboard that visualizes U.S. population data from 2010–2019.

➽ How to Make Proximity Maps with Python: This blog post walks through creating a "distance from" map using Python to calculate distances between universities in the Southeastern Conference (SEC) for college football. It details coding steps to visualize travel distances from one school to others on a contour map, ideal for analyzing team travel or other location-based data.

➽ Data Leakage in Preprocessing: This article addresses data leakage in machine learning, where test data unintentionally influences training data during preprocessing. Common issues include imputing missing values using the mean of the entire dataset, blending test insights into training, which skews model performance.

➽ The Ultimate Guide to RAGs — Each Component Dissected: This blog explores Retrieval Augmented Generation (RAG) in Large Language Models, where relevant data is first retrieved from external sources, then combined with user queries to produce more accurate responses. The RAG approach helps improve accuracy, reduce hallucinations, and provide up-to-date information efficiently.

➽ Beyond Attention: How Advanced Positional Embedding Methods Improve upon the Original Approach in Transformer Architecture. This article explains how the Transformer architecture improved AI models by enabling faster processing and capturing long-range relationships in data through self-attention. Positional embeddings, like sinusoidal and learned encodings, help maintain order, making models work well across different data types.

We’ve got more great things coming your way—see you soon!