POW

For Developers

Setup

Install dependencies

python -m venv venv
source venv/bin/activate
pip install -r requirements/dev.txt
pre-commit install

Download datasets

python -m scripts.download_data --data all

Run POW experiments

• Step localization experiments

Example:

python -m src.experiments.step_localization.evaluate --algorithm=POW --keep_percentile 0.3 --reg 3 --use_unlabeled

Read src/evaluate.py for more details.

• Weizmann classification 1-nn experiments

Example :

python -m src.experiments.weizmann.knn_eval --test_size 0.5 --outlier_ratio 0.1 --metric pow  --m 0.9 --reg 1 --distance euclidean

Read src/experiments/weizmann/knn_eval.py for more details.

• UCR classification k-nn experiments

Example

python -m src.experiments.ucr.knn_eval --dataset=Chinatown --outlier_ratio 0.1 --metric pow  --m 0.9 --reg 1 --distance euclidean --seed 1

• Note : to run softdtw follow instructions in this repo

• Sample notebook for POW and POW with segment regularization in notebook