Continual learning: a feature extraction formalization, an efficient algorithm, and fundamental obstructions
Nov 28, 2022
Speakers
Binghui Peng
Speaker · 0 followers
Andrej Risteski
Speaker · 0 followers
About
Continual learning is an emerging paradigm in machine learning, wherein a model is exposed in an online fashion to data from multiple different distributions (i.e. environments), and is expected to adapt to the distribution change. Precisely, the goal is to perform well in the new environment, while simultaneously retaining the performance on the previous environments (i.e. avoid “catastrophic forgetting”).While this setup has enjoyed a lot of attention in the applied community, there hasn't be theoretical work that even formalizes the desired guarantees. In this paper, we propose a framework for continual learning through the framework of feature extraction—namely, one in which features, as well as a classifier, are being trained with each environment. When the features are linear, we design an efficient gradient-based algorithm 𝖣𝖯𝖦𝗋𝖺𝖽, that is guaranteed to perform well on the current environment, as well as avoid catastrophic forgetting. In the general case, when the features are non-linear, we show such an algorithm cannot exist, whether efficient or not.Continual learning is an emerging paradigm in machine learning, wherein a model is exposed in an online fashion to data from multiple different distributions (i.e. environments), and is expected to adapt to the distribution change. Precisely, the goal is to perform well in the new environment, while simultaneously retaining the performance on the previous environments (i.e. avoid “catastrophic forgetting”).While this setup has enjoyed a lot of attention in the applied community, there hasn't be…
Organizer
NeurIPS 2022
Account · 952 followers
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Fast and Robust Rank Aggregation against Model Misspecification
Watch later
Pan Yuangang, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Imitating Past Successes Can be Very Suboptimal
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Improved Fine-Tuning by Better Leveraging Pre-Training Data
Watch later
Ziquan Liu, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Maximum Class Separation as Inductive Bias in One Matrix
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Improved Deep Neural Network Generalization Using m-Sharpness-Aware Minimization
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Segmenting Moving Objects via an Object-Centric Layered Representation
Watch later
Junyu Xie, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%