{
  "id": "1912.01098",
  "version": "v1",
  "published": "2019-12-02T22:12:16.000Z",
  "updated": "2019-12-02T22:12:16.000Z",
  "title": "Using Dimensionality Reduction to Optimize t-SNE",
  "authors": [
    "Rikhav Shah",
    "Sandeep Silwal"
  ],
  "comment": "11th Annual Workshop on Optimization for Machine Learning (OPT2019 )",
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "t-SNE is a popular tool for embedding multi-dimensional datasets into two or three dimensions. However, it has a large computational cost, especially when the input data has many dimensions. Many use t-SNE to embed the output of a neural network, which is generally of much lower dimension than the original data. This limits the use of t-SNE in unsupervised scenarios. We propose using \\textit{random} projections to embed high dimensional datasets into relatively few dimensions, and then using t-SNE to obtain a two dimensional embedding. We show that random projections preserve the desirable clustering achieved by t-SNE, while dramatically reducing the runtime of finding the embedding.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-12-02T22:12:16.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "dimensionality reduction",
      "optimize t-sne",
      "large computational cost",
      "embed high dimensional datasets",
      "random projections preserve"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}