machine learning, read from the papers up
every algorithm,
built from scratch
13 modules, 43 papers, numpy only. each page has the paper, the code, and a demo that checks one against the other.
00000000000
0111111111111111110
0011111111111111111111111
01111111111111111111111111110
111111111110 011111111111110
11111111110 01111111111
1111111110 0111111111
1111111111 011111111
01111111110 11111111
1111111110 11111111 0111
01111111110 01111110 1111111110 011111
01111111111111 111111111111111110 111111
1111111111110 01111111111111110 011111
1111111111111 11111111111110 111111
1111111111110 111111110 111111
11111111111110 01111111
11111111111110 011111111
111111111111110 0111111111
11111111111111110 01111111111
11111111111111111100 011111111111
111111111111111111111111111111111111111111
111111111111111111111111111111111111
11111111111111111111111111111111
11111111111111111111111111
1111111111111111
the modules
00
Automatic Differentiation
1 paper · Baydin
open →01
Linear Regression, Ridge and Lasso
2 papers · Hoerl, Tibshirani
open →02
Logistic Regression
1 paper · Cox
open →03
k-Nearest Neighbours
1 paper · Cover
open →04
Naive Bayes
1 paper · McCallum
open →05
Decision Trees
2 papers · Quinlan, Breiman
open →06
Bagging and Random Forests
2 papers · Breiman
open →07
Boosting: AdaBoost and Gradient Boosting
2 papers · Freund, Friedman
open →08
Support Vector Machines
3 papers · Boser, Cortes…
open →09
k-Means Clustering
2 papers · Lloyd, Arthur
open →10
Principal Component Analysis
2 papers · Pearson, Hotelling
open →11
Perceptron, MLP and Backpropagation
2 papers · Rosenblatt, Rumelhart
open →11
Activation Functions
10 papers · LeCun, Bridle…
open →11
Optimizers: Momentum, RMSProp, Adam, AdamW
5 papers · Polyak, Sutskever…
open →11
Training Techniques: Init, Dropout, Batch Norm, Schedules, Early Stopping
7 papers · Glorot, He…
open →12
Convolutional Neural Networks
1 paper · LeCun
open →