Name
Abstract | ||
|---|---|---|
We study Winner-Takes-All and rank based Vector Quanti- zation along the lines of the statistical physics of off-line learning. Typical behavior of the system is obtained within a model where high-dimensional training data are drawn from a mixture of Gaussians. The analysis be- comes exact in the simplifying limit of high training temperature. Our main findings concern the existence of phase transitions, i.e. a critical or discontinuous dependence of VQ performance on the training set size. We show how the nature and properties of the transition depend on the num- ber of prototypes and the control parameter of rank based cost functions. |
Year | Venue | Keywords |
|---|---|---|
2008 | ESANN | phase transition,statistical physics,mixture of gaussians,winner take all,cost function |
Field | DocType | Citations |
Training set,Phase transition,Computer science,Vector quantization,Artificial intelligence,Machine learning,Mixture model | Conference | 0 |
PageRank | References | Authors |
0.34 | 3 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Aree Witoelar | 1 | 27 | 4.45 |
| Anarta Ghosh | 2 | 156 | 13.81 |
| Michael Biehl | 3 | 784 | 62.50 |