Zero-shot generalization across architectures for visual classification
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917656165089280 |
|---|---|
| author | Gerritz, Evan Dyballa, Luciano Zucker, Steven W. |
| author_facet | Gerritz, Evan Dyballa, Luciano Zucker, Steven W. |
| contents | Generalization to unseen data is a key desideratum for deep networks, but its relation to classification accuracy is unclear. Using a minimalist vision dataset and a measure of generalizability, we show that popular networks, from deep convolutional networks (CNNs) to transformers, vary in their power to extrapolate to unseen classes both across layers and across architectures. Accuracy is not a good predictor of generalizability, and generalization varies non-monotonically with layer depth. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_14095 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Zero-shot generalization across architectures for visual classification Gerritz, Evan Dyballa, Luciano Zucker, Steven W. Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning I.2.6; I.5.1; I.4.10 Generalization to unseen data is a key desideratum for deep networks, but its relation to classification accuracy is unclear. Using a minimalist vision dataset and a measure of generalizability, we show that popular networks, from deep convolutional networks (CNNs) to transformers, vary in their power to extrapolate to unseen classes both across layers and across architectures. Accuracy is not a good predictor of generalizability, and generalization varies non-monotonically with layer depth. |
| title | Zero-shot generalization across architectures for visual classification |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning I.2.6; I.5.1; I.4.10 |
| url | https://arxiv.org/abs/2402.14095 |