machine-learning-visualized

Why we included this project

Machine Learning Visualized is a Jupyter Book built around a simple idea: you learn an algorithm better when you can watch it work. Each notebook builds an algorithm from the ground up in NumPy, works through the math step by step, and ends with a visualization of the training run converging on its optimal weights, covering gradient descent, perceptrons, logistic regression, k-means, PCA, autoencoders, and neural networks. The interactive Marimo notebooks go further, letting you adjust weights and watch how the loss function responds, which builds intuition for concepts like loss landscapes and backpropagation. The repository is the build configuration for the book, pulling notebooks from separate per-algorithm repos, so it is a learning resource and reference rather than something to deploy. Students, self-taught engineers, and anyone preparing for interviews or grad school will find a clear visual path from optimization basics to neural network internals.