Information Maximization and Language Acquisition
ICANN '01 Proceedings of the International Conference on Artificial Neural Networks
Information-Theoretic Competitive Learning with Inverse Euclidean Distance Output Units
Neural Processing Letters
Automatic inference of cabinet approval ratings by information-theoretic competitive learning
ICONIP'06 Proceedings of the 13th international conference on Neural Information Processing - Volume Part II
| Hi-index | 0.00 |
We have previously described an unsupervised learning procedurethat discovers spatially coherent properties of the world bymaximizing the information that parameters extracted from differentparts of the sensory input convey about some common underlyingcause. When given random dot stereograms of curved surfaces, thisprocedure learns to extract surface depth because that is theproperty that is coherent across space. It also learns how tointerpolate the depth at one location from the depths at nearbylocations (Becker and Hinton 1992b). In this paper, we propose twonew models that handle surfaces with discontinuities. The firstmodel attempts to detect cases of discontinuities and reject them.The second model develops a mixture of expert interpolators. Itlearns to detect the locations of discontinuities and to invokespecialized, asymmetric interpolators that do not cross thediscontinuities.