continual-learning

Continual Learning

Regularization-based

1. 2017 - PNAS - EWC - Overcoming catastrophic forgetting in neural networks [paper] [code] [Fisher Information Matrix] [Natural Gradient Descent] [On Quadratic Penalties in Elastic Weight Consolidation]
2. 2017 - ICML - SI - Continual Learning Through Synaptic Intelligence [paper] [code]
3. 2018 - ECCV - MAS - Memory Aware Synapses: Learning what (not) to forget [paper] [code]
4. 2018 - ICLR - VCL - VARIATIONAL CONTINUAL LEARNING [paper] [code]
5. 2019 - NIPS - UCL - Uncertainty-based Continual Learning with Adaptive Regularization [paper] [code]
6. 2021 - NIPS - AGS-CL - Continual Learning with Node-Importance based Adaptive Group Sparse Regularization [paper]
7. 2016 - ECCV - LwF - Learning without Forgetting [paper] [code]

Architecture-based

1. 2018 - CVPR - PackNet - PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning [paper] [code]
2. 2018 - PMLR - HAT - Overcoming catastrophic forgetting with hard attention to the task [paper] [code]
3. 2018 - ICML - SpaceNet - SpaceNet: Make Free Space For Continual Learning [paper] [code]
4. 2016 - DeepMind - PNN - Progressive Neural Networks [paper] [code] [code]
5. 2018 - ICLR - DEN - Lifelong Learning with Dynamically Expandable Networks [paper] [code]
6. 2019 - NIPS - CPG - Compacting, Picking and Growing for Unforgetting Continual Learning [paper] [code]
7. 2019 - ICML - LtG - Learn to Grow: A Continual Structure Learning Framework for Overcoming Catastrophic Forgetting [paper]
8. 2018 - PiggyBack - Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights [paper] [code]
9. 2017 - PathNet - PathNet: Evolution Channels Gradient Descent in Super Neural Networks [paper]
10. 2017 - DAN - Incremental Learning Through Deep Adaptation [paper]

Memory-based

1. 2017 - CVPR - iCaRL - iCaRL: Incremental Classifier and Representation Learning [paper] [code]
2. 2019 - CVPR - BiC - Large Scale Incremental Learning [paper] [code]
3. 2020 - CVPR - Mnemonics Training: Multi-Class Incremental Learning without Forgetting [paper] [code]
4. 2017 - NIPS - Deep Generative Replay - Continual Learning with Deep Generative Replay [paper] [code]
5. 2019 - ICLR - PGMA - OVERCOMING CATASTROPHIC FORGETTING FOR CONTINUAL LEARNING VIA MODEL ADAPTATION [paper] [code]
6. 2019 - CVPR - DGM - Learning to Remember: A Synaptic Plasticity Driven Framework for Continual Learning [paper] [code]
7. 2020 - NIPS - MERLIN - Meta-Consolidation for Continual Learning [paper] [code]
8. 2020 - CVPR - Generative Feature Replay - Generative Feature Replay For Class-Incremental Learning [paper] [code]

Gradient-based

1. 2020 - PMLR - OGD - Orthogonal gradient descent for continual learning [paper]
2. 2017 - NIPS - GEM - Gradient episodic memory for continual learning [paper] [code]
3. 2019 - ICLR - A-GEM - Efficient lifelong learning with a-gem [paper] [code]
4. 2020 - SOFT-GEM - Gradient Episodic Memory with a Soft Constraint for Continual Learning [paper]
5. 2021 - CVPR - G-Decomposition - Layerwise optimization by gradient decomposition for continual learning [paper]
6. 2020 - NIPS - MEGA - Improved schemes for episodic memory-based lifelong learning [paper] [code]
7. 2021 - ICLR - GPM - Gradient projection memory for continual learning [paper] [code]
8. 2020 - NIPS - ORTHO - Continual learning in low-rank orthogonal subspaces [paper] [code]
9. 2019 - OWM - Continual learning of context-dependent processing in neural networks [paper] [code]

Variational Inference

Week 2

1. 2011 - NIPS - Practical Variational Inference for Neural Networks [paper]
2. 2015 - ICML - Weight Uncertainty in Neural Networks [paper] [code] [code]
3. 2015 - NIPS - Variational Dropout and the Local Reparameterization Trick [paper]

Week 3

1. 2017 - ICML - Variational Dropout Sparsifies Deep Neural Networks [paper] [code]
2. 2018 - NIPS - Variational Dropout via Empirical Bayes [paper]
3. 2019 - CVPR - Variational Bayesian Dropout with a Hierarchical Prior [paper]

Seminar

1. 2021 - Alleviate Representation Overlapping in Class Incremental Learning by Contrastive Class Concentration [paper]
2. 2021 - ICCV - RECALL: Replay-based Continual Learning in Semantic Segmentation [paper] [code]
3. 2021 - ICCV - Striking a Balance between Stability and Plasticity for Class-Incremental Learning [paper]
4. 2019 - NIPS - Random Path Selection for Incremental Learning [paper] [code]
5. 2021 - NIPS - Optimizing Reusable Knowledge for Continual Learning via Metalearning [paper] [code]
6. 2021 - CVPR - Supervised Contrastive Replay: Revisiting the Nearest Class Mean Classifier in Online Class-Incremental Continual Learning [paper] [code]

Bi-week Seminar

DGM

1. 2020 - ICLR - A critical analysis of self-supervision, or what we can learn from a single image [paper] [code]

CL

1. 2020 - Learning to Continually Learn [paper] [code]