Joint optimization of an autoencoder for clustering and embedding

Publikation: Beiträge in ZeitschriftenZeitschriftenaufsätzeForschungbegutachtet

Authors

Deep embedded clustering has become a dominating approach to unsupervised categorization of objects with deep neural networks. The optimization of the most popular methods alternates between the training of a deep autoencoder and a k-means clustering of the autoencoder’s embedding. The diachronic setting, however, prevents the former to benefit from valuable information acquired by the latter. In this paper, we present an alternative where the autoencoder and the clustering are learned simultaneously. This is achieved by providing novel theoretical insight, where we show that the objective function of a certain class of Gaussian mixture models (GMM’s) can naturally be rephrased as the loss function of a one-hidden layer autoencoder thus inheriting the built-in clustering capabilities of the GMM. That simple neural network, referred to as the clustering module, can be integrated into a deep autoencoder resulting in a deep clustering model able to jointly learn a clustering and an embedding. Experiments confirm the equivalence between the clustering module and Gaussian mixture models. Further evaluations affirm the empirical relevance of our deep architecture as it outperforms related baselines on several data sets.

OriginalspracheEnglisch
ZeitschriftMachine Learning
Jahrgang110
Ausgabenummer7
Seiten (von - bis)1901-1937
Anzahl der Seiten37
ISSN0885-6125
DOIs
PublikationsstatusErschienen - 01.07.2021

Dokumente

DOI

Zuletzt angesehen

Publikationen

  1. Using haar wavelets for fault detection in technical processes
  2. Some model properties to control a permanent magnet machine using a controlled invariant subspace
  3. Improving forest ecosystem functions by optimizing tree species spatial arrangement
  4. Impulsive Feedback Linearization for Decoupling of a Constant Disturbance with Low Relative Degree to Control Maglev Systems
  5. Optimizing price levels in e-commerce applications
  6. Direct parameter specification of an attention shift: Evidence from perceptual latency priming
  7. How can employment relations in global value networks be managed towards social responsibility?
  8. Using smart services as a key enabler for collaboration in global production networks
  9. Structure and Organization of Product Development Projects
  10. CaO dissolution during melting and solidification of a Mg-10 wt.% CaO alloy detected with in situ synchrotron radiation diffraction
  11. Learning linear classifiers sensitive to example dependent and noisy costs
  12. Global patterns of ecologically unequal exchange
  13. Iconography on Scientific Instruments. Introduction
  14. A panel cointegrating rank test with structural breaks and cross-sectional dependence
  15. Information Technology in Environmental Engineering
  16. Frames of systems change in sustainability transformations: Lessons from sociotechnical systems and circular economy case studies
  17. Gläserne Bienen (1957)
  18. The efficiency of German public theaters: a stochastic frontier analysis approach
  19. Assessing pre-travel online destination experience values of destination websites
  20. Towards a global understanding of tree mortality
  21. Erkenntnistheorie
  22. Global trait–environment relationships of plant communities
  23. Investigating values and environmental attitudes in the context of the COVID-19 pandemic
  24. Beating uncontrolled eating
  25. Technology-centred learning processes as digital artistic development
  26. Non-target Analysis and Chemometric Evaluation of a Passive Sampler Monitoring of Small Streams
  27. Round, just-below, or precise prices? Cultural differences in the prevalence of price endings in E-commerce