Bayesian Continual Learning

Bayesian continual learning aims to develop machine learning systems that can learn new tasks sequentially without forgetting previously acquired knowledge, a crucial challenge for building truly adaptable AI. Current research focuses on improving the robustness of Bayesian methods, exploring architectures like Bayesian neural networks and spiking neural networks, and addressing issues like model misspecification and data imbalance through techniques such as Elastic Weight Consolidation. This field is significant because it addresses a fundamental limitation of current AI, paving the way for more efficient and adaptable systems with improved uncertainty quantification, applicable to diverse domains like robotics and personalized medicine.

Papers

January 4, 2023

On Sequential Bayesian Inference for Continual Learning
Samuel Kessler, Adam Cobb, Tim G. J. Rudner, Stefan Zohren, Stephen J. Roberts
Continual LEArning Bayesian Neural Network Sequential Bayesian Inference Bayesian Continual Learning

October 28, 2022

Elastic Weight Consolidation Improves the Robustness of Self-Supervised Learning Methods under Transfer
Andrius Ovsianas, Jason Ramapuram, Dan Busbridge, Eeshan Gunesh Dhekane, Russ Webb
Native Robustness Self Supervised Formality Transfer Self Supervised Representation Learning Fine Tuned Model Self Supervised Learning Method Weight Consolidation Bayesian Continual Learning

August 29, 2022

Bayesian Continual Learning via Spiking Neural Networks
Nicolas Skatchkovsky, Hyeryung Jang, Osvaldo Simeone
Neural Network Neuromorphic Computing Continual Adaptation Synaptic Weight Neuromorphic Engineering Bayesian Continual Learning

Bayesian Continual Learning

Papers

On Sequential Bayesian Inference for Continual Learning

Elastic Weight Consolidation Improves the Robustness of Self-Supervised Learning Methods under Transfer

Bayesian Continual Learning via Spiking Neural Networks