Machine Learning Algorithms

This repository contains all my implemented machine learning algorithms to understand their theoretical framework.

K Nearest Neighbor Image Classifier

We build a k nearest neighbor image classifier from scratch to understand its underlying working mechanisms. Then we train and test its performance on CIFAR-10 image dataset.

The following is the cross validation on k graph (cross validation accuracy against k).

Linear Regression

This folder contains a detailed and comprehensive introduction to Linear Regression model, including verifying its assumptions. We use it on Kaggle house prices for prediction.

Neural Networks

We build a two-layer neural network from scratch using solely numpy library to reduce running time by using vectorization. By using cross-validation to tune our hyperparameters, we use them on test set to determine their performace.

The following plot recorded all underlying weights images for us to understand what the neurons in hidden layer do to input images.

SVM Image Classifier

We build a multi-class SVM Image Classifier from scratch using solely numpy library to reduce running time by using vectorization. Then we test its performance on CIFAR-10 image dataset