The SVM Classifier.

This is an implementation of the SVM classifier on the MNIST dataset

Use this resource for a brief introduction
Use this resource for implementation details in matlab

Setup Instructions

This implementation is developed in python 3.9.15 Setup your python environment by following the instructions

With virtualenv

1. Change to the project directory. (the next commands require you to be here)
2. run virtualenv env --python=<path to python 3.9 executable>
3. run source ./env/bin/activate
4. run the bash script pip_installs or follow the last 2 steps. (they do the same thing)
5. run pip install -r requirements.txt
6. run pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu116

With Anaconda

1. change to the project directory. (the env.yml file is in the project root)
2. run conda env create -f env.yml

Usage Instructions

Using the environment built from the instructions above, start your jupyter notebook server by calling
jupyter notebook in the project root.

Then you can follow the jupyter notebook svm_example.ipynb to train a model, generate graphs from training, and then visualize the weights of the model