Learning to Remember: A Synaptic Plasticity Driven Framework for Continual Learning

Oleksiy Ostapenko, Mihai Puscas, Tassilo Klein, Patrick Jähnichen, Moin Nabi

2019 · 27 citations · 17 references

DOIFull text

Open access

Concepts

TL;DR

Continual learning models must acquire knowledge from an unbounded data stream while preserving past information and scaling to growing data volumes. This work proposes Dynamic Generative Memory, a synaptic‑plasticity driven framework for continual learning. DGM employs conditional GANs with neural‑masking of layer activations or connection weights, dynamically expands network capacity guided by learned binary masks, and is evaluated on class‑incremental visual classification.

Abstract

Models trained in the context of continual learning (CL) should be able to learn from a stream of data over an undefined period of time. The main challenges herein are: 1) maintaining old knowledge while simultaneously benefiting from it when learning new tasks, and 2) guaranteeing model scalability with a growing amount of data to learn from. In order to tackle these challenges, we introduce Dynamic Generative Memory (DGM) - synaptic plasticity driven framework for continual learning. DGM relies on conditional generative adversarial networks with learnable connection plasticity realized with neural masking. Specifically, we evaluate two variants of neural masking: applied to (i) layer activations and (ii) to connection weights directly. Furthermore, we propose a dynamic network expansion mechanism that ensures sufficient model capacity to accommodate for continually incoming tasks. The amount of added capacity is determined dynamically from the learned binary mask. We evaluate DGM in the continual class-incremental setup on visual classification tasks.

References

17