/TimeSeriesExplainability

Review of methods for ml explainability on time series data

Primary LanguageJupyter NotebookMIT LicenseMIT

TimeSeriesExplainability

Review of methods for time-series explainability of Neural Networks.

The demo presented here uses pretrained models for speech-based age group classification.

Installation

In order to install, run the following chain of commands: 0. First, get the submodules

git submodule update --init --recursive
  1. Create python virtualenv and install the requirements:
python3 -m virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
  1. Inside the virtualnevironment install the ftdnn submodule:
./install.sh

Overview

For all the details about the algorithms used please check the demo notebook time_series_xplain.ipynb. There is little point in copy-pasting its contents here.