Verified65% confidenceFactExact time
Elastic Weight Consolidation (EWC) estimates the importance of parameters to old tasks via the Fisher information matrix and applies stronger regularization penalties to important parameters to mitigate catastrophic forgetting.
3
Sources
65%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe mainstream approaches to addressing catastrophic forgetting fall into three categories: parameter isolation methods (such as Progressive Neural Networks), regularization-based methods (such as the EWC elastic weight consolidation algorithm), and experience replay methods.73% similarUnverifiedWhen capacity is insufficient to accommodate two tasks, EWC can only slow down forgetting without fundamentally resolving the capacity bottleneck, and the overhead increases as the number of tasks grows due to the accumulation of constraint terms.67% similarUnverifiedEWC在新任务训练时增加惩罚项L_new + λ·Σ_i F_i·(θ_i - θ*_i)²，Fisher信息值越高的参数惩罚力度越大，越难被修改66% similarUnverifiedFine-tuning carries the risk of catastrophic forgetting, where fine-tuning for a specific behavior may degrade the model's performance on other tasks.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/796827API
curl https://kongchang.com/api/v1/knowledge/claims/796827MCP
get_claim(id=796827)