{
  "id": "1805.05814",
  "version": "v1",
  "published": "2018-05-14T14:21:23.000Z",
  "updated": "2018-05-14T14:21:23.000Z",
  "title": "SHADE: Information-Based Regularization for Deep Learning",
  "authors": [
    "Michael Blot",
    "Thomas Robert",
    "Nicolas Thome",
    "Matthieu Cord"
  ],
  "comment": "IEEE International Conference on Image Processing (ICIP) 2018. arXiv admin note: substantial text overlap with arXiv:1804.10988",
  "categories": [
    "stat.ML",
    "cs.LG"
  ],
  "abstract": "Regularization is a big issue for training deep neural networks. In this paper, we propose a new information-theory-based regularization scheme named SHADE for SHAnnon DEcay. The originality of the approach is to define a prior based on conditional entropy, which explicitly decouples the learning of invariant representations in the regularizer and the learning of correlations between inputs and labels in the data fitting term. Our second contribution is to derive a stochastic version of the regularizer compatible with deep learning, resulting in a tractable training scheme. We empirically validate the efficiency of our approach to improve classification performances compared to standard regularization schemes on several standard architectures.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2018-05-14T14:21:23.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "deep learning",
      "information-based regularization",
      "information-theory-based regularization scheme named shade",
      "training deep neural networks",
      "standard regularization schemes"
    ],
    "tags": [
      "conference paper"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}