Factorial Learning by Clustering Features

Abstract

We introduce a novel algorithm for factorial learning, motivated by segmentation problems in computational vision, in which the underlying factors correspond to clusters of highly correlated input features. The algorithm derives from a new kind of competitive clustering model, in which the cluster generators compete to ex(cid:173) plain each feature of the data set and cooperate to explain each input example, rather than competing for examples and cooper(cid:173) ating on features, as in traditional clustering algorithms. A natu(cid:173) ral extension of the algorithm recovers hierarchical models of data generated from multiple unknown categories, each with a differ(cid:173) ent, multiple causal structure. Several simulations demonstrate the power of this approach.

Cite

Text

Tenenbaum and Todorov. "Factorial Learning by Clustering Features." Neural Information Processing Systems, 1994.

Markdown

[Tenenbaum and Todorov. "Factorial Learning by Clustering Features." Neural Information Processing Systems, 1994.](https://mlanthology.org/neurips/1994/tenenbaum1994neurips-factorial/)

BibTeX

@inproceedings{tenenbaum1994neurips-factorial,
  title     = {{Factorial Learning by Clustering Features}},
  author    = {Tenenbaum, Joshua B. and Todorov, Emanuel V.},
  booktitle = {Neural Information Processing Systems},
  year      = {1994},
  pages     = {561-568},
  url       = {https://mlanthology.org/neurips/1994/tenenbaum1994neurips-factorial/}
}