Is deep-learning-from-scratch open source?

Yes — oreilly-japan/deep-learning-from-scratch is open source, released under the MIT license.

What language is deep-learning-from-scratch written in?

oreilly-japan/deep-learning-from-scratch is primarily written in Jupyter Notebook.

How popular is deep-learning-from-scratch?

oreilly-japan/deep-learning-from-scratch has 4.8k stars on GitHub.

Where can I find deep-learning-from-scratch?

oreilly-japan/deep-learning-from-scratch is on GitHub at https://github.com/oreilly-japan/deep-learning-from-scratch.

← all repositories

oreilly-japan/deep-learning-from-scratch

Neural networks with nothing but NumPy and patience

Companion code for a Japanese O'Reilly book that teaches deep learning by building neural networks from scratch using only NumPy and Matplotlib.

★4.8k stars Jupyter Notebook Learning ML Frameworks

View on GitHub ↗

Not currently ranked — collecting fresh signals.

star history

What it does This repository holds the companion code for the Japanese O’Reilly book Deep Learning from Scratch (2016). It collects chapter-by-chapter Python implementations of neural networks, backpropagation, and optimization, relying solely on NumPy and Matplotlib. The files are split into eight chapter folders, shared common utilities, and dataset loaders.

The interesting bit The project treats high-level frameworks as a distraction until you understand the calculus underneath. With no TensorFlow or PyTorch in the dependency list, every forward pass and gradient update is written out as explicit matrix operations.

Key highlights

Implements core concepts from scratch using only NumPy and Matplotlib
Organized into eight chapter folders (ch01 through ch08) plus common and dataset
Provides Jupyter notebooks ready for Amazon SageMaker Studio Lab
MIT licensed, with a public errata wiki
4,799 stars and a widely cited Japanese-language learning reference

Caveats

The README and book are entirely in Japanese; non-Japanese readers will need external context
Code explanations live in the printed book, so the repository alone is essentially uncommented sample code
Published in 2016, so the techniques reflect that era’s pedagogical style

Verdict A useful stop for developers who want to see backpropagation unmasked as raw matrix math. Skip it if you need production-grade framework recipes or English-language documentation.

Frequently asked

What is oreilly-japan/deep-learning-from-scratch?: Companion code for a Japanese O'Reilly book that teaches deep learning by building neural networks from scratch using only NumPy and Matplotlib.
Is deep-learning-from-scratch open source?: Yes — oreilly-japan/deep-learning-from-scratch is open source, released under the MIT license.
What language is deep-learning-from-scratch written in?: oreilly-japan/deep-learning-from-scratch is primarily written in Jupyter Notebook.
How popular is deep-learning-from-scratch?: oreilly-japan/deep-learning-from-scratch has 4.8k stars on GitHub.
Where can I find deep-learning-from-scratch?: oreilly-japan/deep-learning-from-scratch is on GitHub at https://github.com/oreilly-japan/deep-learning-from-scratch.