ProgLearn

https://zenodo.org/badge/DOI/10.5281/zenodo.4060264.svg https://circleci.com/gh/neurodata/ProgLearn/tree/main.svg?style=shield&circle-token=:circle-token https://codecov.io/gh/neurodata/ProgLearn/branches/main/graph/badge.svg https://img.shields.io/pypi/v/proglearn.svg https://img.shields.io/badge/arXiv-2004.12908-red.svg?style=flat https://img.shields.io/badge/License-MIT-blue https://img.shields.io/netlify/97f86f49-81ed-4292-a100-f7031b54ecc7 https://img.shields.io/pypi/dm/proglearn.svg

ProgLearn (Progressive Learning) is a package for exploring and using omnidirectional learning algorithms proposed in our paper (https://arxiv.org/abs/2004.12908).

Motivation

In biological learning, data are used to improve performance simultaneously on the current task, as well as previously encountered and as yet unencountered tasks. In contrast, classical machine learning starts from a blank slate, or tabula rasa, using data only for the single task at hand. While typical transfer learning algorithms can improve performance on future tasks, their performance on prior tasks degrades upon learning new tasks (called "catastrophic forgetting"). Many recent approaches have attempted to maintain performance given new tasks. But striving to avoid forgetting sets the goal unnecessarily low: the goal of ProgLearn is to improve performance on all tasks (including past and future) with any new data.

Python

Python is a powerful programming language that allows concise expressions of network algorithms. Python has a vibrant and growing ecosystem of packages that ProgLearn uses to provide more features such as numerical linear algebra and plotting. In order to make the most out of ProgLearn you will want to know how to write basic programs in Python. Among the many guides to Python, we recommend the Python documentation.

Free software

ProgLearn is free software; you can redistribute it and/or modify it under the terms of the MIT. We welcome contributions. Join us on GitHub.

History

ProgLearn was founded in February 2020. The source code was designed, written and verified by Will LeVine, Jayanta Dey, and Hayden Helm. The experiment code was designed and written by Jayanta Dey and Will LeVine. The repository is maintained by Jayanta Dey and Haoyin Xu.

Indices and tables