2016

Log-time and Log-space Extreme Classification

Jasinska, Kalina, Karampatziakis, Nikos

Understand

We present LTLS, a technique for multiclass and multilabel prediction that can perform training and inference in logarithmic time and space.

  • LTLS embeds large classification problems into simple structured prediction problems and relies on efficient dynamic programming algorithms for inference.
  • We train LTLS with stochastic gradient descent on a number of multiclass and multilabel datasets and show that despite its small memory footprint it is often competitive with existing approaches.

Built on

  • A family of additive online algorithms for category ranking

    Koby Crammer and Yoram Singer · 2003

    Earlier work this paper cites.

  • Wsabie: Scaling up to large vocabulary image annotation

    J. Weston, S. Bengio, and N. Usunier · 2011

    Earlier work this paper cites.

  • FastXML: A fast, accurate and stable tree-classifier for extreme multi-label learning

    Y. Prabhu and M. Varma · 2014

    Earlier work this paper cites.

Similar

  • Sparse local embeddings for extreme multi-label classification

    K. Bhatia, H. Jain, P. Kar, M. Varma, and Jain P · 2015

    Cited alongside, same era.

  • Logarithmic time online multiclass prediction

    A. Choromanska and J. Langford · 2015

    Cited alongside, same era.

  • Fast label embeddings via randomized linear algebra

    P. Mineiro and N. Karampatziakis · 2015

    Cited alongside, same era.

Then

  • Large-scale multi-label learning with missing labels

    H. Yu, P. Jain, P. Kar, and I. Dhillon · 2015

    Later among the works it cites.

  • Pd-sparse : A primal and dual sparse approach to extreme multiclass and multilabel classification

    I. En-Hsu Yen, X. Huang, P. Ravikumar, K. Zhong, and I. Dhillon · 2016

    Closest in time.

  • Extreme f-measure maximization using sparse probability estimates

    K. Jasinska, K. Dembczynski, R. Busa-Fekete, K. Pfannschmidt, T. Klerx, and E. Hullermeier · 2016

    Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…