Information-Theoretic Progress Measures reveal Grokking is an Emergent Phase Transition
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911992873222144 |
|---|---|
| author | Clauw, Kenzo Stramaglia, Sebastiano Marinazzo, Daniele |
| author_facet | Clauw, Kenzo Stramaglia, Sebastiano Marinazzo, Daniele |
| contents | This paper studies emergent phenomena in neural networks by focusing on grokking where models suddenly generalize after delayed memorization. To understand this phase transition, we utilize higher-order mutual information to analyze the collective behavior (synergy) and shared properties (redundancy) between neurons during training. We identify distinct phases before grokking allowing us to anticipate when it occurs. We attribute grokking to an emergent phase transition caused by the synergistic interactions between neurons as a whole. We show that weight decay and weight initialization can enhance the emergent phase. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_08944 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Information-Theoretic Progress Measures reveal Grokking is an Emergent Phase Transition Clauw, Kenzo Stramaglia, Sebastiano Marinazzo, Daniele Machine Learning Information Theory This paper studies emergent phenomena in neural networks by focusing on grokking where models suddenly generalize after delayed memorization. To understand this phase transition, we utilize higher-order mutual information to analyze the collective behavior (synergy) and shared properties (redundancy) between neurons during training. We identify distinct phases before grokking allowing us to anticipate when it occurs. We attribute grokking to an emergent phase transition caused by the synergistic interactions between neurons as a whole. We show that weight decay and weight initialization can enhance the emergent phase. |
| title | Information-Theoretic Progress Measures reveal Grokking is an Emergent Phase Transition |
| topic | Machine Learning Information Theory |
| url | https://arxiv.org/abs/2408.08944 |