التفاصيل البيبلوغرافية
| العنوان: |
Model performance on ImageNet and V1 neural predictivity. |
| المؤلفون: |
Nathan C. L. Kong (11919978), Eshed Margalit (11919981), Justin L. Gardner (8538303), Anthony M. Norcia (8555189) |
| سنة النشر: |
2022 |
| المجموعة: |
Smithsonian Institution: Digital Repository |
| مصطلحات موضوعية: |
Cell Biology, Neuroscience, Physiology, Biotechnology, Ecology, Environmental Sciences not elsewhere classified, Biological Sciences not elsewhere classified, Mathematical Sciences not elsewhere classified, Information Systems not elsewhere classified, ventral visual stream, spatial frequency tuning, spatial frequency preference, power law exponents, power law exponent, lower spatial frequencies, one would indicate, fine stimulus features, decay slightly faster, macaque v1 cells, +task%22">xlink "> task, show striking similarities, quantitatively better models, predictivity , make incorrect predictions, imperceptible image perturbations, macaque v1 eigenspectrum, image perturbations, slow decay, least one, extract features |
| الوصف: |
This table lists all the models we used, their macaque V1 neural response predictivity and their top-1 accuracies on ImageNet validation set images which have not been perturbed (i.e., for a perturbation, δ , and for any norm, ‖ δ ‖ = 0) or were adversarially perturbed with different norm constraints on the perturbations: ‖ δ ‖ ∞ ≤ 1/1020, ‖ δ ‖ 2 ≤ 0.15, ‖ δ ‖ 1 ≤ 40. For the models trained to be adversarially robust, the suffix corresponds to the norm constraint imposed on the size of the perturbation during model training. For example, robust_resnet50_l2_3 corresponds to a ResNet-50 adversarially trained to be robust to perturbations, δ , of size at most ‖ δ ‖ 2 ≤ 3 [ 28 ], igr_robust_resnet50 corresponds to a ResNet-50 trained with input gradient regularization (IGR, [ 21 ]) and resnet50_simclr corresponds to a ResNet-50 trained with the SimCLR unsupervised loss function [ 47 ]. (PDF) |
| نوع الوثيقة: |
article in journal/newspaper |
| اللغة: |
unknown |
| Relation: |
https://figshare.com/articles/journal_contribution/Model_performance_on_ImageNet_and_V1_neural_predictivity_/18053037 |
| DOI: |
10.1371/journal.pcbi.1009739.s008 |
| الاتاحة: |
https://doi.org/10.1371/journal.pcbi.1009739.s008 |
| Rights: |
CC BY 4.0 |
| رقم الانضمام: |
edsbas.F0E32ADF |
| قاعدة البيانات: |
BASE |