Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation

📰 ArXiv cs.AI

Learn how catastrophic forgetting in continual adaptation can be understood through a function-space theory, helping to identify vulnerable output-space directions and predict forgetting vectors

advanced Published 17 Jun 2026
Action Steps
  1. Apply the function-space theory to identify vulnerable output-space directions in your model
  2. Use the cross-task kernel to predict the forgetting vector before new-task training
  3. Analyze the NTK regime to understand the effects of new-task training on old-task prediction
  4. Implement a frozen-backbone linear classifier to test the theory
  5. Evaluate the performance of your model using the closed-form predictor for the forgetting vector
Who Needs to Know This

ML researchers and engineers working on continual learning and adaptation tasks can benefit from this theory to improve model performance and reduce forgetting

Key Insight

💡 Catastrophic forgetting can be understood and predicted through a function-space theory, allowing for more effective continual adaptation

Share This
💡 Catastrophic forgetting is low-rank! New function-space theory helps identify vulnerable directions & predict forgetting vectors #continuallearning #ML

Key Takeaways

Learn how catastrophic forgetting in continual adaptation can be understood through a function-space theory, helping to identify vulnerable output-space directions and predict forgetting vectors

Full Article

Title: Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation

Abstract:
arXiv:2606.18024v1 Announce Type: cross Abstract: Catastrophic forgetting in continual adaptation is usually studied through parameter drift, replay, or distillation, but these views do not identify which output-space directions are vulnerable. We give a function-space account in the NTK regime: new-task training induces old-task prediction drift through the cross-task kernel, yielding a closed-form predictor for the forgetting vector before any new-task gradient step. In frozen-backbone linear-
Read full paper → ← Back to Reads

Related Videos

SQLite3 Tutorial - Learn SQL for Python in 17 Minutes
SQLite3 Tutorial - Learn SQL for Python in 17 Minutes
Thomas Janssen
How to Train AI to Play Games ? How AI Learns to Play ? Several Methods EXPLAINED
How to Train AI to Play Games ? How AI Learns to Play ? Several Methods EXPLAINED
MaxonShire
Introduction to Machine Learning: Lesson 05
Introduction to Machine Learning: Lesson 05
Stephen Blum
Pytorch Embedding Model Part 1
Pytorch Embedding Model Part 1
Stephen Blum
Introduction to Machine Learning: Lesson 04
Introduction to Machine Learning: Lesson 04
Stephen Blum
Introduction to Machine Learning: Lesson 03
Introduction to Machine Learning: Lesson 03
Stephen Blum