Machine Learning Q and AI: 30 Essential Questions and Answers on Machine Learning and AI

Description

Cover Page Title Page Copyright Page Dedication Page About the Author About the Technical Reviewer BRIEF CONTENTS CONTENTS IN DETAIL FOREWORD ACKNOWLEDGMENTS INTRODUCTION Who Is This Book For? What Will You Get Out of This Book? How to Read This Book Online Resources PART I: NEURAL NETWORKS AND DEEP LEARNING 1. EMBEDDINGS, LATENT SPACE, AND REPRESENTATIONS Embeddings Latent Space Representation Exercises References 2. SELF-SUPERVISED LEARNING Self-Supervised Learning vs. Transfer Learning Leveraging Unlabeled Data Self-Prediction and Contrastive Self-Supervised Learning Exercises References 3. FEW-SHOT LEARNING Datasets and Terminology Exercises 4. THE LOTTERY TICKET HYPOTHESIS The Lottery Ticket Training Procedure Practical Implications and Limitations Exercises References 5. REDUCING OVERFITTING WITH DATA Common Methods Collecting More Data Data Augmentation Pretraining Other Methods Exercises References 6. REDUCING OVERFITTING WITH MODEL MODIFICATIONS Common Methods Regularization Smaller Models Caveats with Smaller Models Ensemble Methods Other Methods Choosing a Regularization Technique Exercises References 7. MULTI-GPU TRAINING PARADIGMS The Training Paradigms Model Parallelism Data Parallelism Tensor Parallelism Pipeline Parallelism Sequence Parallelism Recommendations Exercises References 8. THE SUCCESS OF TRANSFORMERS The Attention Mechanism Pretraining via Self-Supervised Learning Large Numbers of Parameters Easy Parallelization Exercises References 9. GENERATIVE AI MODELS Generative vs. Discriminative Modeling Types of Deep Generative Models Energy-Based Models Variational Autoencoders Generative Adversarial Networks Flow-Based Models Autoregressive Models Diffusion Models Consistency Models Recommendations Exercises References 10. SOURCES OF RANDOMNESS Model Weight Initialization Dataset Sampling and Shuffling Nondeterministic Algorithms Different Runtime Algorithms Hardware and Drivers Randomness and Generative AI Exercises References PART II: COMPUTER VISION 11. CALCULATING THE NUMBER OF PARAMETERS How to Find Parameter Counts Convolutional Layers Fully Connected Layers Practical Applications Exercises 12. FULLY CONNECTED AND CONVOLUTIONAL LAYERS When the Kernel and Input Sizes Are Equal When the Kernel Size Is 1 Recommendations Exercises 13. LARGE TRAINING SETS FOR VISION TRANSFORMERS Inductive Biases in CNNs ViTs Can Outperform CNNs Inductive Biases in ViTs Recommendations Exercises References PART III: NATURAL LANGUAGE PROCESSING 14. THE DISTRIBUTIONAL HYPOTHESIS Word2vec, BERT, and GPT Does the Hypothesis Hold? Exercises References 15. DATA AUGMENTATION FOR TEXT Synonym Replacement Word Deletion Word Position Swapping Sentence Shuffling Noise Injection Back Translation Synthetic Data Recommendations Exercises References 16. SELF-ATTENTION Attention in RNNs The Self-Attention Mechanism Exercises References 17. ENCODER- AND DECODER-STYLE TRANSFORMERS The Original Transformer Encoders Decoders Encoder-Decoder Hybrids Terminology Contemporary Transformer Models Exercises References 18. USING AND FINE-TUNING PRETRAINED TRANSFORMERS Using Transformers for Classification Tasks In-Context Learning, Indexing, and Prompt Tuning Parameter-Efficient Fine-Tuning Reinforcement Learning with Human Feedback Adapting Pretrained Language Models Exercises References 19. EVALUATING GENERATIVE LARGE LANGUAGE MODELS Evaluation Metrics for LLMs Perplexity BLEU Score ROUGE Score BERTScore Surrogate Metrics Exercises References PART IV: PRODUCTION AND DEPLOYMENT 20. STATELESS AND STATEFUL TRAINING Stateless (Re)training Stateful Training Exercises 21. DATA-CENTRIC AI Data-Centric vs. Model-Centric AI Recommendations Exercises References 22. SPEEDING UP INFERENCE Parallelization Vectorization Loop Tiling Operator Fusion Quantization Exercises References 23. DATA DISTRIBUTION SHIFTS Covariate Shift Label Shift Concept Drift Domain Shift Types of Data Distribution Shifts Exercises References PART V: PREDICTIVE PERFORMANCE AND MODEL EVALUATION 24. POISSON AND ORDINAL REGRESSION Exercises 25. CONFIDENCE INTERVALS Defining Confidence Intervals The Methods Method 1: Normal Approximation Intervals Method 2: Bootstrapping Training Sets Method 3: Bootstrapping Test Set Predictions Method 4: Retraining Models with Different Random Seeds Recommendations Exercises References 26. CONFIDENCE INTERVALS VS. CONFORMAL PREDICTIONS Confidence Intervals and Prediction Intervals Prediction Intervals and Conformal Predictions Prediction Regions, Intervals, and Sets Computing Conformal Predictions A Conformal Prediction Example The Benefits of Conformal Predictions Recommendations Exercises References 27. PROPER METRICS The Criteria The Mean Squared Error The Cross-Entropy Loss Exercises 28. THE K IN K-FOLD CROSS-VALIDATION Trade-offs in Selecting Values for k Determining Appropriate Values for k Exercises References 29. TRAINING AND TEST SET DISCORDANCE Exercises 30. LIMITED LABELED DATA Improving Model Performance with Limited Labeled Data Labeling More Data Bootstrapping the Data Transfer Learning Self-Supervised Learning Active Learning Few-Shot Learning Meta-Learning Weakly Supervised Learning Semi-Supervised Learning Self-Training Multi-Task Learning Multimodal Learning Inductive Biases Recommendations Exercises References AFTERWORD APPENDIX: ANSWERS TO THE EXERCISES Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Chapter 7 Chapter 8 Chapter 9 Chapter 10 Chapter 11 Chapter 12 Chapter 13 Chapter 14 Chapter 15 Chapter 16 Chapter 17 Chapter 18 Chapter 19 Chapter 20 Chapter 21 Chapter 22 Chapter 23 Chapter 24 Chapter 25 Chapter 26 Chapter 27 Chapter 28 Chapter 29 Chapter 30 INDEX