Human Centered Machine Learning

Getting started

Create python 3.10 environment:

python3 -m venv venv
source venv/bin/activate

Install dependencies:

pip install -r requirements.txt

Train model

python src/train.py --model mlp --dataset mnist --label_noise 0.1

Get baseline

Note that the arguments for this script are the same as for train.py

python src/calculate_baseline.py --model mlp --dataset mnist --epochs 10 --label_noise 0.1

Get influences

Note that the arguments for this script are the same as for train.py

python src/get_influences.py --model mlp --dataset mnist --epochs 10 --label_noise 0.1

Get evaluation on noise prediction

Note that the arguments for this script are the same as for train.py, this also saves a plot of the noise prediction

python src/evaluate_noise_prediction.py --model mlp --dataset mnist --epochs 10 --label_noises 0.1 0.2 0.3 0.4
python src/evaluate_noise_prediction.py --model resnet18 --dataset isic2024 --epochs 50 --label_noises 0.1 0.2 0.3 0.4
python src/evaluate_noise_prediction.py --model resnet50 --dataset isic2024 --epochs 50 --label_noises 0.1 0.2 0.3 0.4

Retrain model

python src/train.py --model mlp --dataset mnist --label_noise 0.1

Analyse

python src/plot.py --data ..

python src/get_op_pp.py --data ...

Snellius

Install enviroment; run job