Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Neocognitron capable of incremental learning
Fukushima K. Neural Networks17 (1):37-46,2004.Type:Article
Date Reviewed: Apr 22 2004

The learning algorithms within neocognitron are discussed in this paper. Neocognitron is a neural network model that has two distinctive types of cells: S-cells and C-cells. The prime use of this type of neural network is in the imitation of the visual cortex system. The paper does not present any comparisons between neocognitron-type neural networks and other models used for pattern recognition, such as Hopfield neural nets.

The paper presents neocognitron learning mechanisms that have incremental learning capabilities. The author proposes a learning method and network architecture. The algorithms and the network behavior were tested through a computer simulation examining the scale of the network, and the recognition rate.

The analysis of the experimental results is somewhat limited, and does not clearly demonstrate the benefits of the new incremental capabilities. The author might consider doing more experiments in addition to the limited number of experiments presented here. The list of references was also rather short. In addition, the author made reference to his own work four times, in 14 listed references.

Reviewer:  Aladdin Ayesh Review #: CR129492 (0410-1222)
Bookmark and Share
  Reviewer Selected
Connectionism And Neural Nets (I.2.6 ... )
Analysis Of Algorithms (I.1.2 ... )
Concept Learning (I.2.6 ... )
Algorithms (I.1.2 )
Learning (I.2.6 )
Would you recommend this review?
Other reviews under "Connectionism And Neural Nets": Date
Connectionist robot motion planning
Mel B., Academic Press Prof., Inc., San Diego, CA, 1990. Type: Book (9780124900202)
Apr 1 1992
Foundations of neural networks
Khanna T., Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1990. Type: Book (9780201500363)
Jan 1 1993
Hecht-Nielsen R., Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1989. Type: Book (9780201093551)
Jan 1 1993

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2023 ThinkLoud®
Terms of Use
| Privacy Policy