Welcome to the new version of CaltechAUTHORS. Login is currently restricted to library staff. If you notice any issues, please email coda@library.caltech.edu
Published September 2007 | Published
Book Section - Chapter Open

Unsupervised Category Discovery in Images Using Sparse Neural Coding

Abstract

We present an unsupervised method for learning and recognizing object categories from unlabeled images. Motivated by the existence of highly selective, sparsely firing cells observed in the human medial temporal lobe (MTL), we apply a sparse generative model to the outputs of a biologically faithful model of the primate ventral visual system. In our model, a network of nonlinear neurons learns a sparse representation of its inputs through an unsupervised expectation-maximization process. In recognition, this model is used in a maximum-likelihood manner to classify unseen images, and we find units emerging from learning that respond selectively to specific image categories. A significant advantage of this approach is that there is no need to specify the number of categories present in the training set. We present classification accuracy using three different evaluation metrics.

Additional Information

We thank Thomas Serre and Minjoon Kouh of MIT for providing the visual system model used here as well as assistance with its operation, and Richard Murray, Jerry Marsden, and Pietro Perona at Caltech and Bruno Olshausen at Berkeley for valuable feedback. This work was funded by a Fannie and John Hertz Foundation Fellowship (to S.W.), as well as by grants from the ONR, NIMH, NSF, and DARPA.

Attached Files

Published - paper-144.pdf

Files

paper-144.pdf
Files (1.1 MB)
Name Size Download all
md5:f51d26a5da98086af3c04ab00576fe33
1.1 MB Preview Download

Additional details

Created:
September 15, 2023
Modified:
October 23, 2023