Roadmap

Team mission

Develop a domain agnostic deep learning for time series forecasting, classification, and anomaly detection framework and leverage it solve pressing problems in health, climate, agriculture, and other high impact humanitarian areas.

Project information

General Framework Enhancements
Machine Learning Research Papers/Sprints
COVID-19 County Dashboard

Roadmap overview

As explained above we have two major focuses: (1) general repository enhancements that can be applied to any time series problem and (2) vertical impact focus areas that can aim to tackle a specific issue (e.g. COVID spread, flash flood prediction…).

Core Repo

COVID-19

Climate

Feature 1

Feature 2

Meta-Data Model Enhancements

Benchmarking

Adding new models

COVID Model Finalization

COVID Dashboard Development

Feature 3

River flow dataset V1

Detailed quarterly roadmap

More

Core Repository Focus Areas

Feature	Priority	Effort	Status	Notes

Feature	Priority	Effort	Status	Notes
Adding new time series models/loss functions	Medium	Medium	In progress	Leverage PyData Global Sprint
Meta-data incorporation	HIGH	HIGH	In progress
Model Interpretability	HIGH	Medium	In progress	Should add methods other than SHAP.
Increasing test coverage	HIGH	Medium	In progress	Need more unit tests of models and meta-data.
Documentation and tutorials	Medium	Medium	In progress
Auto experimentation	Low	HIGH	not STARTED	This would be useful but requires a lot of effort.
Cloud Provider Integration	Medium	Medium	In progress
Multitask learning support	Medium	HIGH	not STARTED

Application/Impact Focus Areas

Feature	Priority	Effort	Status	Notes

Feature	Priority	Effort	Status	Notes
COVID County Dashboard	HIGH	HIGH	In progress	Link to Slideshow
River flow dashboard	mEdium	HIGH	NOT STARTED
River flow open-source	mEdium	HIGH	IN PROGRESS

Research Focus Areas

Focus Area	Priority	Effort	Status	Notes

Focus Area	Priority	Effort	Status	Notes
Incorporating meta-data in TS models
Transfer learning in time series application

Focus areas