Web-app and backend code for ML visual dashboard The App is built in Dash framework (https://dash.plotly.com) and can be run in any IDE supporting Python.
If you have Conda, simply run
source setup.sh
First time initialisation creates a suitable Conda Python environment with the appropriate dependencies for this App.
To run the ML visual dashboard in a local environment, do
python runApp.py
and copy the provided link into a web browser. It should look something like
http://127.0.0.1:7777
A Dockerfile is provided to allow deployment of this webapp as a Docker image which also indicated the commands needed to run the ML WebApp if one wishes to bypass pakcage installation in conda and use menv instead.
Backend data needed to power the various plots and displays on the UI are generated and stored in csv format. Once RegenerateModels.py has been called once (either through setup.sh or in isolation), the ML visual dashboard can run entirely disconnected from the internet.
The backend data is generated using ATLAS Open Data and using it to train and assess various scikit-learn neural network configurations. The setup is seeded so should lead to reproducible results accross machines.
The Dash framework is a free toolkit for powering web-apps using Python-based templates and functions. The core features are given here. For details relevant to parts of this Web-App, please refer to the following:
-
Caley Yardley, University of Sussex (caley.luce.yardley@cern.ch). Developer & maintainer.
-
Giovanni Guerrieri, CERN (giovanni.guerrieri@cern.ch). Maintainer.
-
Andrey Kukhmay, University of Sussex. Prototype developer.