mini-project-2

Contents

statement - problem statement

report - my write up

code - contains code for the training/validation/testing of four models - Multi-layer Perceptron, Linear Support Vector Machine, Kernel Support Vector Machine, and Logistic regression - all implemented using scikit-learn.

slurm-scripts - contains some simple slurm scripts that were used for running the models on the abacus campus cluster and for obtaining the results, and their outputs.

plots - contains a few plots that were generated for the write up

Requirements

• Python 3.7
• scikit-learn
• numpy

Running the code

cd ./code

# run mlp classifier with predetermined hyperparameters
python mlp --best

# search for best hidden layer dimensions of model
python mlp.py --search_lu 

# search for best learning rate with fixed hidden layer dimensions
python mlp.py --search_lr

# run simple example of overfitting
python mlp.py --overfit

# run mlp classifier after transforming samples with pca
python mlp.py --pca --best

# run mlp classifier after transforming samples with lda
python mlp.py --lda --best

# run mlp classifier after transforming samples with pca followed by lda
python mlp.py --pca --lda --best

# run mlp classifier after scaling samples
python mlp.py --scaling --best

# run mlp classifier after scaling and mean subtracting samples
python mlp.py --mean-sub --scaling --best

The other classifiers can be run similarly by changing the file name and search arguments.