ICLR 2023 accepted papers on Continual learning
27 of the 1,575 papers accepted at ICLR 2023 (1.7%) are on Continual learning, compared with 2.2% at ICLR 2022. They fall into 4 topics.
See ICLR 2023 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Online and replay-based · 16 papers
- Measuring Forgetting of Memorized Training Examples (15 citations)
- Thalamus: a brain-inspired algorithm for biologically-plausible continual learning and disentangled representations (9 citations)
- Continual evaluation for lifelong learning: Identifying the stability gap (7 citations)
- New Insights for the Stability-Plasticity Dilemma in Online Continual Learning (4 citations)
- Sparse Distributed Memory is a Continual Learner (4 citations)
- Task-Aware Information Routing from Common Representation Space in Lifelong Learning (4 citations)
- Building a Subspace of Policies for Scalable Continual Learning (3 citations)
- Artificial Neuronal Ensembles with Learned Context Dependent Gating (2 citations)
- Error Sensitivity Modulation based Experience Replay: Mitigating Abrupt Representation Drift in Continual Learning (1 citation)
- Is Forgetting Less a Good Inductive Bias for Forward Transfer? (1 citation)
- Sequential Learning of Neural Networks for Prequential MDL (1 citation)
- Online Boundary-Free Continual Learning by Scheduled Data Prior
and 4 more in the interactive map.
Class-incremental · 7 papers
- A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning (45 citations)
- Neural Collapse Inspired Feature-Classifier Alignment for Few-Shot Class-Incremental Learning (22 citations)
- On the Soft-Subnetwork for Few-Shot Class Incremental Learning (12 citations)
- Incremental Learning of Structured Memory via Closed-Loop Transcription (2 citations)
- Warping the Space: Weight Space Rotation for Class-Incremental Few-Shot Learning
- BEEF: Bi-Compatible Class-Incremental Learning via Energy-Based Expansion and Fusion
- Progressive Voronoi Diagram Subdivision Enables Accurate Data-free Class-Incremental Learning