Table of contents

1. Installation
2. Sample Usage

Installation

pip install -r requirements.txt

Sample Usage

Classification using all features

Proposed Multi-CNN applied to raw data - superficial vs deep OP (Male) at region 1200cm^-1

python multicnn.py --dataset "./dataset/raw/1200_superficial_vs_deep_OP_M.csv" --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1 --channel pspline_aspls jbcd beads ria fabc adaptive_minmax goldindec;

Baseline CNN applied to raw data - superficial vs deep OP (Male) at region 1200cm^-1

python cnn.py --dataset "./dataset/raw/1200_superficial_vs_deep_OP_M.csv" --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1;

Baseline CNN applied to pre-processed data - superficial vs deep OP (Male) at region 1200cm^-1

python cnn.py --dataset "./dataset/preprocessed/1200_superficial_vs_deep_OP_M.csv" --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1;

Feature selection

Please run the proposed Multi-CNN on the classification task prior to the following experiments:

SVM applied to raw data - superficial vs deep OP (Male) at region 1200cm^-1 using top 5 features

python svm_dt_reduced.py --method "svm" --dataset "./dataset/raw/1200_superficial_vs_deep_OP_M.csv" --num-feat 5 --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1

Decision Tree applied to raw data - superficial vs deep OP (Male) at region 1200cm^-1 using top 5 features

python svm_dt_reduced.py --method "dt" --dataset "./dataset/raw/1200_superficial_vs_deep_OP_M.csv" --num-feat 5 --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1

ANN applied to raw data - superficial vs deep OP (Male) at region 1200cm^-1 using top 5 features

python ann_reduced.py --dataset "./dataset/raw/1200_superficial_vs_deep_OP_M.csv" --num-feat 5 --neurons 22 --seed 0 --threads 1 --lr 0.01 --train-batch 4 --valid-batch 1 --epochs 100 --checkpoint-hist 1;