# Image Classification on Tiny ImageNet (Natural Accuracy and Weight Regularization Metrics)

Source: [Image Classification on Tiny ImageNet (Natural Accuracy and Weight Regularization Metrics)](<https://www.sota2.com/research/sota/image-classification-on-tiny-imagenet-natural-accuracy-and-weight-regularization-metrics>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T06:08:44.926Z

Evaluation of Image Classification performance on the Tiny ImageNet dataset, focusing on Natural Accuracy alongside specific model regularization metrics such as weight norm (\|\|W\|\|2) and associated loss terms.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Image Classification |
| Dataset (including any stated split) | Tiny ImageNet |
| Metric | Weight L2 Norm |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 2 of 2 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Weight L2 Norm](<https://www.sota2.com/research/sota/image-classification-on-tiny-imagenet-natural-accuracy-and-weight-regularization-metrics.md?metric=Weight%20L2%20Norm>): Lower is better; unit: SCALAR.

- [Natural Accuracy](<https://www.sota2.com/research/sota/image-classification-on-tiny-imagenet-natural-accuracy-and-weight-regularization-metrics.md?metric=Natural%20Accuracy>): Higher is better; unit: PERCENT.

- [AutoAtt Score](<https://www.sota2.com/research/sota/image-classification-on-tiny-imagenet-natural-accuracy-and-weight-regularization-metrics.md?metric=AutoAtt%20Score>): Higher is better; unit: SCALAR.

- [CrossEntropy + L2 Loss](<https://www.sota2.com/research/sota/image-classification-on-tiny-imagenet-natural-accuracy-and-weight-regularization-metrics.md?metric=CrossEntropy%20%2B%20L2%20Loss>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Weight L2 Norm | Result date | Report paper |
| --- | --- | --- | --- |
| Adaptive Weight Decay (AWD) | 15.01 | Not recorded | 68fa71551e209131cea508f2 |
| Non-adaptive Weight Decay | 25.62 | Not recorded | 68fa71551e209131cea508f2 |

### Adaptive Weight Decay (AWD)

Weight L2 Norm: 15.01. Result ID: 6991c5973a6c3148490a7e60.

[Reporting paper](<https://www.sota2.com/research/paper/68fa71551e209131cea508f2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71551e209131cea508f2.md>)

Recorded evaluation configuration: {"lambda\_awd":"0.01714","Optimizer":"SGD"}. This record does not assert independent reproduction.

### Non-adaptive Weight Decay

Weight L2 Norm: 25.62. Result ID: 6991c5973a6c3148490a7e5f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa71551e209131cea508f2.md>)

Recorded evaluation configuration: {"lambda\_wd":"0.00211","Optimizer":"SGD"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.