Practical Ai on the Google Cloud Platform Chorzów

Working with AI is complicated and expensive for many developers. That's why cloud providers have stepped in to make it easier, offering free (or affordable) state-of-the-art models and training tools to get you started. With this book, you'll learn how to use Google's AI-powered cloud services to do everything from creating a chatbot to analyzing text, images, and video.Author Micheal Lanham demonstrates methods for building and training models step-by-step and shows you how to expand your models to accomplish increasingly complex tasks. If you have a good grasp of math and the Python language, you'll quickly get up to speed with Google Cloud Platform, whether you want to build an AI assistant or a simple business AI application.Learn key concepts for data science, machine learning, and deep learningExplore tools like Video AI and AutoML TablesBuild a simple language processor using deep learning systemsPerform image recognition using CNNs, transfer learning, and GANsUse Google's Dialogflow to create chatbots and conversational AIAnalyze video with automatic video indexing, face detection, and TensorFlow HubBuild a complete working AI agent application Spis treści: Preface Who Should Read This Book Why I Wrote This Book Navigating This Book A Note on the Google AI Platform Things You Need for This Book Conventions Used in This Book Using Code Examples OReilly Online Learning How to Contact Us Acknowledgments 1. Data Science and Deep Learning What Is Data Science? Classification and Regression Regression Goodness of Fit Classification with Logistic Regression Multivariant Regression and Classification Data Discovery and Preparation Bad Data Training, Test, and Validation Data Good Data Preparing Data Questioning Your Data The Basics of Deep Learning The Perceptron Game Understanding How Networks Learn Backpropagation Optimization and Gradient Descent Vanishing or Exploding Gradients SGD and Batching Samples Batch Normalization and Regularization Activation Functions Loss Functions Building a Deep Learner Optimizing a Deep Learning Network Overfitting and Underfitting Network Capacity Conclusion Game Answers Game 5 2. AI on the Google Cloud Platform AI Services on GCP The AI Hub AI Platform AI Building Blocks Google Colab Notebooks Building a Regression Model with Colab AutoML Tables The Cloud Shell Managing Cloud Data Conclusion 3. Image Analysis and Recognition on the Cloud Deep Learning with Images Enter Convolution Neural Networks Image Classification Set Up and Load Data Inspecting Image Data Channels and CNN Building the Model Training the AI Fashionista to Discern Fashions Improving Fashionista AI 2.0 Transfer Learning Images Identifying Cats or Dogs Transfer Learning a Keras Application Model Training Transfer Learning Retraining a Better Base Model Object Detection and the Object Detection Hub API YOLO for Object Detection Generating Images with GANs Conclusion 4. Understanding Language on the Cloud Natural Language Processing, with Embeddings Understanding One-Hot Encoding Vocabulary and Bag-of-Words Word Embeddings Understanding and Visualizing Embeddings Recurrent Networks for NLP Recurrent Networks for Memory Classifying Movie Reviews RNN Variations Neural Translation and the Translation API Sequence-to-Sequence Learning Translation API AutoML Translation Natural Language API BERT: Bidirectional Encoder Representations from Transformers Semantic Analysis with BERT Document Matching with BERT BERT for General Text Analysis Conclusion 5. Chatbots and Conversational AI Building Chatbots with Python Developing Goal-Oriented Chatbots with Dialogflow Building Text Transformers Loading and Preparing Data Understanding Attention Masking and the Transformer Encoding and Decoding the Sequence Training Conversational Chatbots Compiling and Training the Model Evaluation and Prediction Using Transformer for Conversational Chatbots Conclusion 6. Video Analysis on the Cloud Downloading Video with Python Video AI and Video Indexing Building a Webcam Face Detector Understanding Face Embeddings Recognizing Actions with TF Hub Exploring the Video Intelligence API Conclusion 7. Generators in the Cloud Unsupervised Learning with Autoencoders Mapping the Latent Space with VAE Generative Adversarial Network Exploring the World of Generators A Path for Exploring GANs Translating Images with Pix2Pix and CycleGAN Translating unpaired images with CycleGAN Attention and the Self-Attention GAN Understanding Self-Attention Self-Attention for Image ColorizationDeOldify Conclusion 8. Building AI Assistants in the Cloud Needing Smarter Agents Introducing Reinforcement Learning Multiarm Bandits and a Single State Choosing the greedy action Adding Quality and Q Learning Playing with OpenAI Gym Exploration Versus Exploitation Balancing exploration and exploitation Understanding Temporal Difference Learning Episodic versus continuous learning On policy versus off policy Model-based versus model-free Discrete versus continuous actions or states Building an Example Agent with Expected SARSA Using SARSA to Drive a Taxi Understanding hierarchical states Learning State Hierarchies with Hierarchical Reinforcement Learning Functional decomposition with MAXQ Bringing Deep to Reinforcement Learning Deep Q Learning Optimizing Policy with Policy Gradient Methods Conclusion 9. Putting AI Assistants to Work Designing an Eat/No Eat AI Selecting and Preparing Data for the AI Training the Nutritionist Model Optimizing Deep Reinforcement Learning Building the Eat/No Eat Agent Testing the AI Agent Commercializing the AI Agent Identifying App/AI Issues Involving Users and Progressing Development Future Considerations Conclusion 10. Commercializing AI The Ethics of Commercializing AI Packaging Up the Eat/No Eat App Reviewing Options for Deployment Deploying to GitHub Deploying with Google Cloud Deploy Exploring the Future of Practical AI Conclusion Index