Name
Abstract | ||
|---|---|---|
We describe a directed bilinear model that learns higher-order groupings among features of natural images. The model represents images in terms of two sets of latent variables: one set of variables represents which feature groups are active, while the other specifies the relative activity within groups. Such a factorized representation is beneficial because it is stable in response to small variations in the placement of features while still preserving information about relative spatial relationships. When trained on MNIST digits, the resulting representation provides state of the art performance in classification using a simple classifier. When trained on natural images, the model learns to group features according to proximity in position, orientation, and scale. The model achieves high log-likelihood (-94 nats), surpassing the current state of the art for natural images achievable with an mcRBM model. |
Year | DOI | Venue |
|---|---|---|
2011 | 10.1109/ICCV.2011.6126473 | ICCV |
Keywords | Field | DocType |
mnist digit,probabilistic model,art performance,mcrbm model,bilinear model,relative spatial relationship,natural image,resulting representation,relative activity,factorized representation,current state,latent variable,higher order,vectors,data models,data model,probability,spatial relationships,mathematical model,computer model,computational modeling,kernel,accuracy,image classification | Data modeling,MNIST database,Computer science,Latent variable,Artificial intelligence,Classifier (linguistics),Contextual image classification,Kernel (linear algebra),Computer vision,Pattern recognition,Statistical model,Machine learning,Bilinear interpolation | Conference |
ISSN | Citations | PageRank |
1550-5499 | 8 | 1.32 |
References | Authors | |
24 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Benjamin J. Culpepper | 1 | 106 | 8.92 |
| Jascha Sohl-Dickstein | 2 | 673 | 82.82 |
| Bruno A. Olshausen | 3 | 493 | 66.79 |