/FilterNet-1

Deep learning architecture for computationally efficient activity recognition as described in the paper FilterNet: A many-to-many deep learning architecture for time series classification

Primary LanguageJupyter Notebook

FilterNet: A many-to-many deep learning architecture for time series classification

DOI

This repository contains code to reproduce the results and figures in the paper: FilterNet: A many-to-many deep learning architecture for time series classification.

Setup

The easiest way to run this software is via the Anaconda Python distribution.

  1. Install Anaconda
  2. Run conda env create -f environment.yaml
  3. Enable the filternet environment, like, source activate filternet
  4. Install filternet so it is importable, by running pip install -e . in the same directory as setup.py

Running tests

In the root dir of this repo:

pytest tests

This will be really slow the first time because it has to download and pre-process several large AR datasets.

Subsequent test runs will probably still be slow, but... less slow.

Reproducing Results

  1. Run the scripts in the scripts/ directory. These are very long-running scripts that reproduce each experimental condition many times. You might want to set, e.g., NUM_REPEATS=1 if you don't need this level of reproducibility.

  2. Run the notebooks to re-produce the figures. You might need to edit a few paths to specific models to match the filenames on your system, especially if you changed the NAME or NUM_REPEATS parameters.

Copyright (C) 2020 Pet Insight Project - All Rights Reserved