{
  "id": "1902.02366",
  "version": "v1",
  "published": "2019-02-06T19:18:13.000Z",
  "updated": "2019-02-06T19:18:13.000Z",
  "title": "Negative eigenvalues of the Hessian in deep neural networks",
  "authors": [
    "Guillaume Alain",
    "Nicolas Le Roux",
    "Pierre-Antoine Manzagol"
  ],
  "categories": [
    "cs.LG",
    "math.OC",
    "stat.ML"
  ],
  "abstract": "The loss function of deep networks is known to be non-convex but the precise nature of this nonconvexity is still an active area of research. In this work, we study the loss landscape of deep networks through the eigendecompositions of their Hessian matrix. In particular, we examine how important the negative eigenvalues are and the benefits one can observe in handling them appropriately.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-02-06T19:18:13.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "deep neural networks",
      "negative eigenvalues",
      "deep networks",
      "loss function",
      "precise nature"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}