Packt+ | Advance your knowledge in tech

You're reading from Artificial Intelligence By Example Develop machine intelligence from scratch using real artificial intelligence use cases

Product type Paperback

Published in May 2018

Publisher Packt

ISBN-13 9781788990547

Length 490 pages

Edition 1st Edition

Languages

Python

Tools

TensorFlow

Concepts

Artificial Intelligence

Author (1):

Denis Rothman

View More author details

Table of Contents (24) Chapters

Title Page

Dedication

Packt Upsell

Contributors

Preface

1. Become an Adaptive Thinker FREE CHAPTER

2. Think like a Machine

3. Apply Machine Thinking to a Human Problem

4. Become an Unconventional Innovator

5. Manage the Power of Machine Learning and Deep Learning

6. Don't Get Lost in Techniques – Focus on Optimizing Your Solutions

7. When and How to Use Artificial Intelligence

8. Revolutions Designed for Some Corporations and Disruptive Innovations for Small to Large Companies

9. Getting Your Neurons to Work

10. Applying Biomimicking to Artificial Intelligence

11. Conceptual Representation Learning

12. Automated Planning and Scheduling

13. AI and the Internet of Things (IoT)

14. Optimizing Blockchains with AI

15. Cognitive NLP Chatbots

16. Improve the Emotional Intelligence Deficiencies of Chatbots

17. Quantum Computers That Think

1. Answers to the Questions

Index

Chapter 7 – When and How to Use Artificial Intelligence

The questions will focus on the hyperparameters.

1. The number of k clusters is not that important. (Yes | No)

The answer is no. The number of clusters requires careful selection, possibly a trial-and-error approach. Each project will lead to different clusters.

2. Mini-batches and batches contain the same amount of data. (Yes | No)

The answer is no. "Batch" generally refers to the dataset and "mini-batch" represents a "subset" of data.

3. K-means can run without mini-batches. (Yes | No)

The answer is yes and no. If the volume of data remains small, then the training epochs can run on the whole dataset. If the data volume exceeds a reasonable amount of computer power (CPU or GPU), mini-batches must be created to optimize training computation.

4. Must centroids be optimized for result acceptance? (Yes | No)

The answer is yes and no. Suppose you want to put a key in a keyhole. The keyhole represents the centroid of your visual cluster. You must be precise. If you are simply throwing a piece of paper in your garbage can, you do not need to aim at the perfect center (centroid) of the cluster (marked by the rim of the garbage can) to attain that goal. Centroid precision depends on what is asked of the algorithm.

5. It does not take long to optimize hyperparameters. (Yes | No)

The answer is yes and no. If it's a simple project, it will not take long. If you are facing a large dataset, it will take some time to find the optimal hyperparameters.

6. Can it sometimes take weeks to train a large dataset? (Yes | No)

The answer is yes. Media hype and hard work are two different worlds. Machine learning and deep learning are are still tough projects to implement.

7. AWS SageMaker only offers a k-means algorithm. (Yes | No)

The answer is no. AWS Sagemaker offers a complete range of algorithms in the job creation interface. You can also use your algorithms choosing Custom.