| Presentation | 1995/6/30 Admissibility of Memorization Learning with respect to Projection Learning Akira Hirabayashi, Hidemitsu Ogawa, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In the training of multilayered feedforward neural networks using the error-backpropagation (BP) algorithm, it has been observed that the error over non-training data may increase, even though the error over training data decreases. This phenomenon is referred to as over-learning. In previous work it was shown how over-learning can be viewed as being the result of using the BP criterion as a substitute for some true criterion. Here, the concept of admissibility, which is defined using the relationship between the two criteria, was introduced. In this paper we consider the case where the true criterion is the projection learning criterion, and give the necessary and sufficient conditions for the projection learning to admit memorization learning in the presense of noise. Based on these conditions, we devised a method of choosing the training sets so that the adimissibility conditions are satisfied. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | multilayered feedforward neural network / generalization ability / over-learning / training sets / admissibility / projection learning |
| Paper # | |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 1995/6/30(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Neurocomputing (NC) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Admissibility of Memorization Learning with respect to Projection Learning |
| Sub Title (in English) | |
| Keyword(1) | multilayered feedforward neural network |
| Keyword(2) | generalization ability |
| Keyword(3) | over-learning |
| Keyword(4) | training sets |
| Keyword(5) | admissibility |
| Keyword(6) | projection learning |
| 1st Author's Name | Akira Hirabayashi |
| 1st Author's Affiliation | Department of Computer Science Graduate School of Information Science and Engineering Tokyo Institute of Technology() |
| 2nd Author's Name | Hidemitsu Ogawa |
| 2nd Author's Affiliation | Department of Computer Science Graduate School of Information Science and Engineering Tokyo Institute of Technology |
| Date | 1995/6/30 |
| Paper # | |
| Volume (vol) | vol.95 |
| Number (no) | 135 |
| Page | pp.pp.- |
| #Pages | 8 |
| Date of Issue | |