{
  "id": "1910.01319",
  "version": "v1",
  "published": "2019-10-03T06:31:58.000Z",
  "updated": "2019-10-03T06:31:58.000Z",
  "title": "An empirical study of pretrained representations for few-shot classification",
  "authors": [
    "Tiago Ramalho",
    "Thierry Sousbie",
    "Stefano Peluchetti"
  ],
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "Recent algorithms with state-of-the-art few-shot classification results start their procedure by computing data features output by a large pretrained model. In this paper we systematically investigate which models provide the best representations for a few-shot image classification task when pretrained on the Imagenet dataset. We test their representations when used as the starting point for different few-shot classification algorithms. We observe that models trained on a supervised classification task have higher performance than models trained in an unsupervised manner even when transferred to out-of-distribution datasets. Models trained with adversarial robustness transfer better, while having slightly lower accuracy than supervised models.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-10-03T06:31:58.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "empirical study",
      "pretrained representations",
      "state-of-the-art few-shot classification results start",
      "adversarial robustness transfer better",
      "few-shot image classification task"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}