Summary of Reducing Catastrophic Forgetting Of Incremental Learning in the Absence Of Rehearsal Memory with Task-specific Token, by Young Jo Choi et al.
Reducing catastrophic forgetting of incremental learning in the absence of rehearsal memory with task-specific token
by Young Jo Choi, Min Kyoon Yoo, Yu Rang Park
First submitted to arxiv on: 6 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The novel method presented in this paper addresses catastrophic forgetting in deep learning models by preserving previous knowledge without storing past data. Inspired by the architecture of a vision transformer, it uses unique tokens to encapsulate compressed knowledge from each task and generate task-specific embeddings. The approach incorporates a distillation process for efficient interactions after multiple learning steps, optimizing against forgetting. Evaluation metrics include accuracy and backward transfer using a benchmark dataset for different task-incremental learning scenarios. The results show the superiority of this approach over compared methods in terms of both accuracy and lowest backward transfer. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper solves a big problem with artificial intelligence called “forgetting.” When machines learn new things, they often forget what they learned before. To fix this, scientists created a new way for machines to remember what they learned earlier without actually keeping the old information. This is important because it helps keep our data and personal info safe while still letting machines get better at their jobs. |
Keywords
* Artificial intelligence * Deep learning * Distillation * Vision transformer
