Algorithms that Get Old: The Case of Generative Deep Neural Networks

被引:0
作者
Turinici, Gabriel [1 ]
机构
[1] Univ Paris 09, CNRS, CEREMADE, PSL, Paris, France
来源
MACHINE LEARNING, OPTIMIZATION, AND DATA SCIENCE, LOD 2022, PT II | 2023年 / 13811卷
关键词
Variational auto-encoder; Generative adversarial network; Statistical distance; Vector quantization; Deep neural network; Measure compression; STATISTICS;
D O I
10.1007/978-3-031-25891-6_14
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Generative deep neural networks used in machine learning, like the Variational Auto-Encoders (VAE), and Generative Adversarial Networks (GANs) produce new objects each time when asked to do so with the constraint that the new objects remain similar to some list of examples given as input. However, this behavior is unlike that of human artists that change their style as time goes by and seldom return to the style of the initial creations. We investigate a situation where VAEs are used to sample from a probability measure described by some empirical dataset. Based on recent works on Radon-Sobolev statistical distances, we propose a numerical paradigm, to be used in conjunction with a generative algorithm, that satisfies the two following requirements: the objects created do not repeat and evolve to fill the entire target probability distribution.
引用
收藏
页码:179 / 187
页数:9
相关论文
共 16 条
  • [11] EQUIVALENCE OF DISTANCE-BASED AND RKHS-BASED STATISTICS IN HYPOTHESIS TESTING
    Sejdinovic, Dino
    Sriperumbudur, Bharath
    Gretton, Arthur
    Fukumizu, Kenji
    [J]. ANNALS OF STATISTICS, 2013, 41 (05) : 2263 - 2291
  • [12] Quantile functions for multivariate analysis: approaches and applications
    Serfling, R
    [J]. STATISTICA NEERLANDICA, 2002, 56 (02) : 214 - 232
  • [13] Silberschatz A., 2018, Operating System Concepts, Vtenth
  • [14] Energy statistics: A class of statistics based on distances
    Szekely, Gabor J.
    Rizzo, Maria L.
    [J]. JOURNAL OF STATISTICAL PLANNING AND INFERENCE, 2013, 143 (08) : 1249 - 1272
  • [15] Turinici Gabriel, 2022, Zenodo, DOI 10.5281/ZENODO.6671523
  • [16] Radon-Sobolev Variational Auto-Encoders
    Turinici, Gabriel
    [J]. NEURAL NETWORKS, 2021, 141 : 294 - 305