{
  "id": "2312.16060",
  "version": "v1",
  "published": "2023-12-26T14:15:19.000Z",
  "updated": "2023-12-26T14:15:19.000Z",
  "title": "Error-free Training for Artificial Neural Network",
  "authors": [
    "Bo Deng"
  ],
  "comment": "10 pages, 3 figures, Matlab mfiles available for online download",
  "categories": [
    "cs.LG",
    "cs.NE",
    "math.DS"
  ],
  "abstract": "Conventional training methods for artificial neural network (ANN) models never achieve zero error rate systematically for large data. A new training method consists of three steps: first create an auxiliary data from conventionally trained parameters which correspond exactly to a global minimum for the loss function of the cloned data; second create a one-parameter homotopy (hybrid) of the auxiliary data and the original data; and third train the model for the hybrid data iteratively from the auxiliary data end of the homotopy parameter to the original data end while maintaining the zero-error training rate at every iteration. This continuationmethod is guaranteed to converge numerically by a theorem which converts the ANN training problem into a continuation problem for fixed points of a parameterized transformation in the training parameter space to which the Uniform Contraction Mapping Theorem from dynamical systems applies.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2023-12-26T14:15:19.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "artificial neural network",
      "error-free training",
      "achieve zero error rate",
      "auxiliary data end",
      "original data end"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 10,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}