{
  "id": "2108.12604",
  "version": "v1",
  "published": "2021-08-28T08:48:31.000Z",
  "updated": "2021-08-28T08:48:31.000Z",
  "title": "Threshold: Pruning Tool for Densely Connected Convolutional Networks",
  "authors": [
    "Rui-Yang Ju",
    "Ting-Yu Lin",
    "Jen-Shiun Chiang"
  ],
  "comment": "4 pages, 3 figures",
  "categories": [
    "cs.CV"
  ],
  "abstract": "Deep neural networks have made significant progress in the field of computer vision. Recent studies have shown that depth, width and shortcut connections of neural network architectures play a crucial role in their performance. One of the most advanced neural network architectures, DenseNet, has achieved excellent convergence rates through dense connections. However, it still has obvious shortcomings in the usage of amount of memory. In this paper, we introduce a new type of pruning tool, threshold, which refers to the principle of the threshold voltage in MOSFET. This work employs this method to connect blocks of different depths in different ways to reduce the usage of memory. It is denoted as ThresholdNet. We compare ThresholdNet with other different networks for FLOPs and memory usage, and the experiments show that ThresholdNet is 70% less memory than that of the original DenseNet.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2021-08-28T08:48:31.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "densely connected convolutional networks",
      "pruning tool",
      "neural network architectures play",
      "thresholdnet",
      "deep neural networks"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 4,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}