This repo contains a numpy from-scratch implementation of some ML algorithms, initially designed for the MNIST Digit classification task (~3% error, averaged over 20 runs of 5 fold cross-validation)

- [x] kNN

- [x] 3 Layer MLP (ReLU & Softmax activations), Cross-Entropy Loss

- [x] Least Squares

- [x] Winnow

- [x] One-vs-One and One-vs-All Muliticlass Kernel Perceptron

- [x] logistic regression with AdaGrad optimiser

- [ ] SVM (primal + dual)

files: CV.py and helper_functions.py

- [x] random train/test split

- [x] Cross Validation

- [x] Gram Matrix (for polynomial and Gaussian kernels)

- [x] numerical gradient check

TODO: change output type from error rate to prediction