Sequoia: A Software Framework to Unify Continual Learning Research

Abstract

The field of Continual Learning (CL) seeks to develop algorithms that accumulate knowledge and skills over time through interaction with nonstationary environments and data distributions. Measuring progress in CL can be difficult because a plethora of evaluation procedures (settings) and algorithmic solutions (methods) have emerged, each with their own potentially disjoint set of assumptions about the CL problem. In this work, we view each setting as a set of assumptions. We then create a tree-shaped hierarchy of the research settings in CL, in which more general settings become the parents of those with more restrictive assumptions. This makes it possible to use inheritance to share and reuse research, as developing a method for a given setting also makes it directly applicable onto any of its children. We instantiate this idea as a publicly available software framework called Sequoia, which features a variety of settings from both the Continual Supervised Learning (CSL) and Continual Reinforcement Learning (CRL) domains. Sequoia also includes a growing suite of methods which are easy to extend and customize, in addition to more specialized methods from third-party libraries. We hope that this new paradigm and its first implementation can serve as a foundation for the unification and acceleration of research in CL. You can help us grow the tree by visiting www.github.com/lebrice/Sequoia.

Tutorial - Continual Learning beyond classification

“…Those two settings are also known as respectively class-incremental and instance-incremental [5,6]. The objectives to be optimized may change over time as well [3], as in continual RL [7,8,9].…”

“…Thus, creating diverse benchmarks, as well as approaches that do not critically rely on the assumptions from the default scenario, should be an ongoing effort. This effort should be pushed notably by existing continual learning libraries such as Continuum [12], Avalanche [13] or Sequoia [7].…”

Continual Object Detection: A review of definitions, strategies, and challenges

“…Additionally, Task-Free or Task-Agnostic CL [22,23] represents an additional scenario for when the task labels are not given during either training or testing, which makes it the most challenging scheme. For that, the model does not have any information on task boundaries and still needs to deal with data distribution changes.…”

CORA: Benchmarks, Baselines, and Metrics as a Platform for Continual Reinforcement Learning Agents

“…Our work is conceptually similar to bsuite [48], which curates a collection of toy, diagnostic experiments to evaluate different capabilities of a standard, non-continual RL agent. Concurrent with our work, Sequoia [46] introduces a software framework with baselines, metrics, and evaluations aimed at unifying research in continual supervised learning and continual reinforcement learning. In contrast to both, we present CORA, a platform that focuses on the continual RL setting and introduces challenging task sequences as benchmarks toward improving different aspects of continual RL agents.…”