ML-From-Scratch

Code uses only NumPy and is written for clarity, making it easy to trace algorithms like backpropagation or gradient descent step-by-step, as shown in the detailed layer summaries.

Implements a wide range from linear regression to GANs and deep Q-networks, providing a holistic learning resource across supervised, unsupervised, and reinforcement learning.

Offers reusable layers like Conv2D and BatchNormalization, allowing users to build custom networks from scratch, as demonstrated in the CNN and GAN examples.

Each implementation comes with runnable examples, such as polynomial regression on temperature data, providing hands-on learning with real-world applications.