reptimeline: Tracking Discrete Representation Evolution During Neural Network Training
Fuente:
Zenodo
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Lingua: | inglese |
| Pubblicazione: |
Zenodo
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866902334591729664 |
|---|---|
| author | Ornelas Brand, J. Arturo |
| author_facet | Ornelas Brand, J. Arturo |
| contents | A Python library for tracking how discrete representations evolve during neural network training. Tracks per-code lifecycle events (births, deaths, connections, phase transitions) and discovers what each code element encodes without requiring prior ontological knowledge. Validated on three backends: binary autoencoder, sparse autoencoder (Pythia-70M), and neurosymbolic projection head. |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_20018087 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | reptimeline: Tracking Discrete Representation Evolution During Neural Network Training Ornelas Brand, J. Arturo representation learning discrete representations interpretability lifecycle tracking phase transitions training dynamics sparse autoencoder VQ-VAE A Python library for tracking how discrete representations evolve during neural network training. Tracks per-code lifecycle events (births, deaths, connections, phase transitions) and discovers what each code element encodes without requiring prior ontological knowledge. Validated on three backends: binary autoencoder, sparse autoencoder (Pythia-70M), and neurosymbolic projection head. |
| title | reptimeline: Tracking Discrete Representation Evolution During Neural Network Training |
| topic | representation learning discrete representations interpretability lifecycle tracking phase transitions training dynamics sparse autoencoder VQ-VAE |
| url | https://doi.org/10.5281/zenodo.20018087 |