Timezone: »
We introduce and study the problem of Online Continual Compression, where one attempts to simultaneously learn to compress and store a representative dataset from a non i.i.d data stream, while only observing each sample once. A naive application of auto-encoder in this setting encounters a major challenge: representations derived from earlier encoder states must be usable by later decoder states. We show how to use discrete auto-encoders to effectively address this challenge and introduce Adaptive Quantization Modules (AQM) to control variation in the compression ability of the module at any given stage of learning. This enables selecting an appropriate compression for incoming samples, while taking into account overall memory constraints and current progress of the learned compression. Unlike previous methods, our approach does not require any pretraining, even on challenging datasets. We show that using AQM to replace standard episodic memory in continual learning settings leads to significant gains on continual learning benchmarks with images, LiDAR, and reinforcement learning agents.
Author Information
Lucas Caccia (McGIll)
Eugene Belilovsky (Mila)
Massimo Caccia (MILA)
Joelle Pineau (McGill University / Facebook)
More from the Same Authors
-
2022 : Building a Subspace of Policies for Scalable Continual Learning »
Jean-Baptiste Gaya · Thang Doan · Lucas Caccia · Laure Soulier · Ludovic Denoyer · Roberta Raileanu -
2023 : Preventing Dimensional Collapse in Contrastive Local Learning with Subsampling »
Louis Fournier · Adeetya Patel · Michael Eickenberg · Edouard Oyallon · Eugene Belilovsky -
2023 : Learning to Optimize with Recurrent Hierarchical Transformers »
Abhinav Moudgil · Boris Knyazev · Guillaume Lajoie · Eugene Belilovsky -
2023 : Re-Weighted Softmax Cross-Entropy to Control Forgetting in Federated Learning »
Gwen Legate · Lucas Caccia · Eugene Belilovsky -
2023 : Guiding The Last Layer in Federated Learning with Pre-Trained Models »
Gwen Legate · Nicolas Bernier · Lucas Caccia · Edouard Oyallon · Eugene Belilovsky -
2023 : Continual Pre-Training of Large Language Models: How to re-warm your model? »
Kshitij Gupta · Benjamin Thérien · Adam Ibrahim · Mats Richter · Quentin Anthony · Eugene Belilovsky · Timothée Lesort · Irina Rish -
2023 Workshop: Localized Learning: Decentralized Model Updates via Non-Global Objectives »
David I. Inouye · Mengye Ren · Mateusz Malinowski · Michael Eickenberg · Gao Huang · Eugene Belilovsky -
2023 : Fostering Women's Leadership in the Realm of Emerging Trends and Technologies »
Joelle Pineau · Rihab Gorsane · Pascale FUNG -
2023 : Joelle Pineau - A culture of open and reproducible research, in the era of large AI generative models »
Joelle Pineau -
2023 Panel: The Societal Impacts of AI »
Sanmi Koyejo · Samy Bengio · Ashia Wilson · Kirikowhai Mikaere · Joelle Pineau -
2023 Poster: Prototype-Sample Relation Distillation: Towards Replay-Free Continual Learning »
Nader Asadi · MohammadReza Davari · Sudhir Mudur · Rahaf Aljundi · Eugene Belilovsky -
2023 Poster: Can Forward Gradient Match Backpropagation? »
Louis Fournier · Stéphane Rivaud SORBONNE UNIVERSITE ISIR · Eugene Belilovsky · Michael Eickenberg · Edouard Oyallon -
2022 Poster: Towards Scaling Difference Target Propagation by Learning Backprop Targets »
Maxence ERNOULT · Fabrice Normandin · Abhinav Moudgil · Sean Spinney · Eugene Belilovsky · Irina Rish · Blake Richards · Yoshua Bengio -
2022 Spotlight: Towards Scaling Difference Target Propagation by Learning Backprop Targets »
Maxence ERNOULT · Fabrice Normandin · Abhinav Moudgil · Sean Spinney · Eugene Belilovsky · Irina Rish · Blake Richards · Yoshua Bengio -
2021 Workshop: ICML 2021 Workshop on Unsupervised Reinforcement Learning »
Feryal Behbahani · Joelle Pineau · Lerrel Pinto · Roberta Raileanu · Aravind Srinivas · Denis Yarats · Amy Zhang -
2021 Poster: OptiDICE: Offline Policy Optimization via Stationary Distribution Correction Estimation »
Jongmin Lee · Wonseok Jeon · Byung-Jun Lee · Joelle Pineau · Kee-Eung Kim -
2021 Spotlight: OptiDICE: Offline Policy Optimization via Stationary Distribution Correction Estimation »
Jongmin Lee · Wonseok Jeon · Byung-Jun Lee · Joelle Pineau · Kee-Eung Kim -
2020 Workshop: Workshop on Continual Learning »
Haytham Fayek · Arslan Chaudhry · David Lopez-Paz · Eugene Belilovsky · Jonathan Richard Schwarz · Marc Pickett · Rahaf Aljundi · Sayna Ebrahimi · Razvan Pascanu · Puneet Dokania -
2020 Workshop: MLRetrospectives: A Venue for Self-Reflection in ML Research »
Jessica Forde · Jesse Dodge · Mayoore Jaiswal · Rosanne Liu · Ryan Lowe · Rosanne Liu · Joelle Pineau · Yoshua Bengio -
2020 Poster: Constrained Markov Decision Processes via Backward Value Functions »
Harsh Satija · Philip Amortila · Joelle Pineau -
2020 Poster: Interference and Generalization in Temporal Difference Learning »
Emmanuel Bengio · Joelle Pineau · Doina Precup -
2020 Poster: Decoupled Greedy Learning of CNNs »
Eugene Belilovsky · Michael Eickenberg · Edouard Oyallon -
2020 Poster: Invariant Causal Prediction for Block MDPs »
Amy Zhang · Clare Lyle · Shagun Sodhani · Angelos Filos · Marta Kwiatkowska · Joelle Pineau · Yarin Gal · Doina Precup -
2019 Workshop: Generative Modeling and Model-Based Reasoning for Robotics and AI »
Aravind Rajeswaran · Emanuel Todorov · Igor Mordatch · William Agnew · Amy Zhang · Joelle Pineau · Michael Chang · Dumitru Erhan · Sergey Levine · Kimberly Stachenfeld · Marvin Zhang -
2019 Poster: Greedy Layerwise Learning Can Scale To ImageNet »
Eugene Belilovsky · Michael Eickenberg · Edouard Oyallon -
2019 Oral: Greedy Layerwise Learning Can Scale To ImageNet »
Eugene Belilovsky · Michael Eickenberg · Edouard Oyallon -
2019 Poster: Separable value functions across time-scales »
Joshua Romoff · Peter Henderson · Ahmed Touati · Yann Ollivier · Joelle Pineau · Emma Brunskill -
2019 Oral: Separable value functions across time-scales »
Joshua Romoff · Peter Henderson · Ahmed Touati · Yann Ollivier · Joelle Pineau · Emma Brunskill -
2018 Poster: Focused Hierarchical RNNs for Conditional Sequence Processing »
Rosemary Nan Ke · Konrad Zolna · Alessandro Sordoni · Zhouhan Lin · Adam Trischler · Yoshua Bengio · Joelle Pineau · Laurent Charlin · Christopher Pal -
2018 Oral: Focused Hierarchical RNNs for Conditional Sequence Processing »
Rosemary Nan Ke · Konrad Zolna · Alessandro Sordoni · Zhouhan Lin · Adam Trischler · Yoshua Bengio · Joelle Pineau · Laurent Charlin · Christopher Pal -
2018 Poster: An Inference-Based Policy Gradient Method for Learning Options »
Matthew Smith · Herke van Hoof · Joelle Pineau -
2018 Oral: An Inference-Based Policy Gradient Method for Learning Options »
Matthew Smith · Herke van Hoof · Joelle Pineau -
2017 Workshop: Reproducibility in Machine Learning Research »
Rosemary Nan Ke · Anirudh Goyal · Alex Lamb · Joelle Pineau · Samy Bengio · Yoshua Bengio -
2017 : Lifelong Learning - Panel Discussion »
Sergey Levine · Joelle Pineau · Balaraman Ravindran · Andrei A Rusu -
2017 : Joelle Pineau: A few modest insights from my lifelong learning »
Joelle Pineau