Description
Deep Learning with Python, Third Edition, authored by François Chollet and Matthew Watson, is a bestselling resource that delves into the intricacies of deep learning using Python. This edition has been updated to include the latest advancements in the field, such as generative AI, Keras 3, PyTorch, and JAX. The book is designed for both beginners and experienced practitioners, offering a clear and practical approach to understanding deep learning concepts and applications.
The book begins by explaining what makes deep learning a transformative approach in artificial intelligence. It highlights how deep learning automates feature engineering, scales efficiently with hardware, and enables the creation of versatile, reusable models that can be adapted to various tasks. Readers will learn about the impact of foundation models, which are trained on extensive datasets using self-supervision and can be fine-tuned for a wide range of downstream tasks, thus reducing the need for task-specific models.
A significant portion of the book is dedicated to the universal workflow for solving machine learning problems. This workflow involves defining the task, developing a model, and deploying it to production while continuously monitoring and improving its performance. The authors provide insights on framing real-world problems as machine learning tasks by understanding the business context, identifying available data, and mapping the problem to suitable machine learning tasks such as classification, regression, or ranking.
The book also covers best practices for designing convolutional neural network architectures, including the use of residual connections, batch normalization, and depthwise separable convolutions. Additionally, it discusses how Vision Transformers differ from traditional convolutional networks in computer vision, emphasizing their ability to process images as sequences of patches and capture long-range dependencies.
Modern object detection models like YOLO and RetinaNet are explored, explaining how they predict bounding boxes and class probabilities in a single step for fast and accurate object detection. The authors also delve into deep learning techniques for time series forecasting and how the Transformer architecture enhances advanced language modeling through attention mechanisms. Finally, strategies for generating coherent text with language models and methods for scaling deep learning training for large models and datasets are discussed, making this book a vital resource for anyone interested in the field of deep learning.
Book Details
Author: François Chollet and Matthew Watson
Publication Year: 2023
Pages: 400
Formats Available: Paperback, eBook
Publisher: Manning Publications
Skill Level: Intermediate to Advanced
Who This Book Is For
- Image Classification
- Object Detection
- Text Generation
- Time Series Forecasting
- Image Segmentation
