Combinatorial optimization by using a neural network operating in block-sequential mode
✍ Scribed by Fumiyuki Shiratani; Kimiaki Yamamoto
- Publisher
- John Wiley and Sons
- Year
- 1994
- Tongue
- English
- Weight
- 490 KB
- Volume
- 25
- Category
- Article
- ISSN
- 0882-1666
No coin nor oath required. For personal study only.
✦ Synopsis
Abstract
In a laterally connected neural network which has continuous output functions and has its state updated in block‐sequential mode at discrete times, a condition inequality is derived for the slope of the output function and the stepwidth of the updating rule which ensures that the network energy decreases monotonically with the state variation. This makes it possible to update in block‐sequential mode, which is more general than synchronous updating or asynchronous updating, application to an arbitrary optimization problem in which the evaluation function is given in a quadratic form. As one example, the method is applied to the traveling salesman problem. The advantages of applying this method as opposed to synchronous and asynchronous updating are discussed, and its effectiveness is confirmed.