English · 简体中文
A hands-on, project-driven guide to modern deep learning — from linear regression all the way to GANs and deep reinforcement learning, with every model paired with runnable code.
10 chapters · ~49 runnable notebooks · Computer Vision · NLP · Finance · Reinforcement Learning
Deep Learning in Action with Python (《Python 深度学习与项目实战》) is a deep learning book published by Posts & Telecom Press (人民邮电出版社) in 2021, part of its Deep Learning Series.
Across 10 chapters, it builds up the core families of modern deep learning step by step — theory paired with code — and puts each one to work on real projects spanning computer vision, natural language processing, finance, and reinforcement learning. Everything is built on Python, TensorFlow, and Keras.
This repository hosts the complete source code for the book: every example and project, organised chapter by chapter as ready-to-run Jupyter notebooks.
📝 Note — The book and the in-notebook comments are written in Chinese. The code itself is standard Python and reads the same in any language.
The book takes you from first principles to advanced, research-grade architectures:
- 🧱 Foundations — gradient descent, linear & logistic regression, Softmax classifiers, fully connected networks, and the building blocks of training (activations, initialisation, loss functions).
- 🎛️ Training that works — fighting overfitting, batch normalisation, and the Keras functional API.
- 👁️ Computer Vision — MNIST & CIFAR-10 classification, a cats-vs-dogs project, classic CNN architectures, and transfer learning.
- 💬 Natural Language Processing — tokenization, word embeddings (including Chinese word vectors), RNN / LSTM / GRU, bidirectional LSTM, attention, ELMo, and text generation.
- 💳 Finance — credit-card fraud detection and stock-price prediction with LSTMs.
- 🎨 Generative & RL — autoencoders, GANs, and deep reinforcement learning with DQN, Policy Gradient, and Actor-Critic.
| # | Topic | Hands-on highlights | Code |
|---|---|---|---|
| 01 | Linear Regression | Gradient descent · Boston housing dataset | 📁 notebooks |
| 02 | Logistic Regression | Binary classification project | 📁 notebooks |
| 03 | Softmax Classifier | Multi-class classification · data preprocessing | 📁 notebooks |
| 04 | Fully Connected Neural Networks | Activations · initialisation · loss functions · MNIST | 📁 notebooks |
| 05 | Optimizing Neural Networks | Regularisation · batch norm · Keras functional API · CIFAR-10 | 📁 notebooks |
| 06 | Convolutional Neural Networks | Conv & pooling layers · cats-vs-dogs · classic CNNs · transfer learning | 📁 notebooks |
| 07 | Recurrent Neural Networks | Embeddings · LSTM / GRU · BiLSTM · attention · ELMo · text generation · stock prediction | 📁 notebooks |
| 08 | Autoencoders | Dimensionality reduction · fraud detection · deconvolution · image denoising | 📁 notebooks |
| 09 | Generative Adversarial Networks | Image generation with GANs | 📁 notebooks |
| 10 | Deep Reinforcement Learning | DQN · Policy Gradient · Actor-Critic | 📁 notebooks |
All code is provided as .ipynb notebooks — open, edit, and run them in Jupyter Notebook or JupyterLab.
# 1. Clone the repository
git clone https://github.com/sagebei/deep_learning_in_action_with_python.git
cd deep_learning_in_action_with_python
# 2. Install the core dependencies
pip install tensorflow keras jupyter numpy pandas matplotlib scikit-learn
# 3. Launch Jupyter and open any chapter
jupyter notebook💡 Datasets are loaded from local paths. Depending on where you store each dataset, you may need to adjust the file path at the top of a notebook so it points to the right location.
📦 Print edition
📱 E-book & read online
Written by Bei Zhou (周北) · github.com/sagebei
If this repository or the book helped you, consider leaving a ⭐ — it helps others find it.
© 2021 Bei Zhou · Published by Posts & Telecom Press (人民邮电出版社) · ISBN 978-7-115-55083-5.
Source code is provided for readers of the book.