ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1903.04056
19
1
v1v2 (latest)

One-Pass Sparsified Gaussian Mixtures

10 March 2019
E. Kightley
Stephen Becker
ArXiv (abs)PDFHTML
Abstract

We present a one-pass sparsified Gaussian mixture model (SGMM). Given NNN data points in PPP dimensions, XXX, the model fits KKK Gaussian distributions to XXX and (softly) classifies each point to these clusters. After paying an up-front cost of O(NPlog⁡P)\mathcal{O}(NP\log P)O(NPlogP) to precondition the data, we subsample QQQ entries of each data point and discard the full PPP-dimensional data. SGMM operates in O(KNQ)\mathcal{O}(KNQ)O(KNQ) time per iteration for diagonal or spherical covariances, independent of PPP, while estimating the model parameters in the full PPP-dimensional space, making it one-pass and hence suitable for streaming data. We derive the maximum likelihood estimators for the parameters in the sparsified regime, demonstrate clustering on synthetic and real data, and show that SGMM is faster than GMM while preserving accuracy.

View on arXiv
Comments on this paper