Enhancement of kohonen's self-organizing map algorithm / Barcenas, Jennylou R. and De Villa, Mary Rose M. 6

By: Barcenas, Jennylou R. and De Villa, Mary Rose M. 4 0 16, [, ] | [, ] |
Contributor(s): 5 6 [] |
Language: Unknown language code Summary language: Unknown language code Original language: Unknown language code Series: ; 200346Edition: Description: 28 cm. 25 ppContent type: text Media type: unmediated Carrier type: volumeISBN: ISSN: 2Other title: 6 []Uniform titles: | | Related works: 1 40 6 []Subject(s): -- 2 -- 0 -- -- | -- 2 -- 0 -- 6 -- | 2 0 -- | -- -- 20 -- | | -- -- -- -- 20 -- | -- -- -- 20 -- --Genre/Form: -- 2 -- Additional physical formats: DDC classification: | LOC classification: | | 2Other classification:
Contents:
Action note: In: Summary: Abstract: This study has the purpose of reducing the current limitations of Kohonen's Self-Organizing Feature Map Algorithm that is mainly used in applications such as robotics, speech, image and pattern recognition, and other artificial-intelligence-related implementation. The main problem of KSOM Algorithm is, it suffers from relative slowness and inefficiently relying on a random shuffling to find the best weight vector to represent an input vector. In the Kohonen network, if weights are to be initialized randomly, most likely, they will not match the input distribution. During training, some weight vectors will be so far away from any input vector that they will never be trained and are hence wasted. To resolved the limitation, the weight vectors should be distributed according to the density of the input vectors. Another problem discussed on this research is that every KSOM is different and finds different similarities among the sample vector. KSOM's organize sample data so that in the final product, the samples are usually surrounded by similar samples. However, similar samples are not always near each other. So a lot of maps are to be constructed in order to get one final good map . And increasing the efficiency in organizing similar sample data will resolve the problem. Another aspect discussed is KSOMs are very computationally expensive which is a major drawback since as the dimensions of the data increases, dimension reduction visualization techniques become more important, but then the time to compute them also increases. The more neighbors use to calculate the distance, the better similarity map will be obtained, but the number of distances the algorithm needed to compute increases exponentially. The proponents will produce the most favorable solution regarding this problem. The proponents will produce the most favorable solution regarding this problem. The proponents performed changes to the existing algorithm to overcome its limitations. Adjusting the neighborhood function and changing some of the parameters concerning the generation and visualization of the feature map will be applied. The proponents want to present useful information to the readers of this research to make them understand some helpful facts and details regarding neutral networks and artificial intelligence. Other editions:
Tags from this library: No tags from this library for this title. Log in to add tags.
    Average rating: 0.0 (0 votes)

Undergraduate Thesis: (Bachelor of Science in Computer Science) - Pamantasan ng Lungsod ng Maynila, 2003. 56

5

Abstract: This study has the purpose of reducing the current limitations of Kohonen's Self-Organizing Feature Map Algorithm that is mainly used in applications such as robotics, speech, image and pattern recognition, and other artificial-intelligence-related implementation. The main problem of KSOM Algorithm is, it suffers from relative slowness and inefficiently relying on a random shuffling to find the best weight vector to represent an input vector. In the Kohonen network, if weights are to be initialized randomly, most likely, they will not match the input distribution. During training, some weight vectors will be so far away from any input vector that they will never be trained and are hence wasted. To resolved the limitation, the weight vectors should be distributed according to the density of the input vectors. Another problem discussed on this research is that every KSOM is different and finds different similarities among the sample vector. KSOM's organize sample data so that in the final product, the samples are usually surrounded by similar samples. However, similar samples are not always near each other. So a lot of maps are to be constructed in order to get one final good map . And increasing the efficiency in organizing similar sample data will resolve the problem. Another aspect discussed is KSOMs are very computationally expensive which is a major drawback since as the dimensions of the data increases, dimension reduction visualization techniques become more important, but then the time to compute them also increases. The more neighbors use to calculate the distance, the better similarity map will be obtained, but the number of distances the algorithm needed to compute increases exponentially. The proponents will produce the most favorable solution regarding this problem. The proponents will produce the most favorable solution regarding this problem. The proponents performed changes to the existing algorithm to overcome its limitations. Adjusting the neighborhood function and changing some of the parameters concerning the generation and visualization of the feature map will be applied. The proponents want to present useful information to the readers of this research to make them understand some helpful facts and details regarding neutral networks and artificial intelligence.

5

There are no comments for this item.

to post a comment.

© Copyright 2024 Phoenix Library Management System - Pinnacle Technologies, Inc. All Rights Reserved.