Limits...
Pattern classification using an olfactory model with PCA feature selection in electronic noses: study and application.

Fu J, Huang C, Xing J, Zheng J - Sensors (Basel) (2012)

Bottom Line: The principal component analysis technique was applied for feature selection and dimension reduction.In the former case the results showed that the average correct classification rate increased as more principal components were put in to feature vector.We concluded that 6~8 channels of the model with principal component feature vector values of at least 90% cumulative variance is adequate for a classification task of 3~5 pattern classes considering the trade-off between time consumption and classification rate.

View Article: PubMed Central - PubMed

Affiliation: College of Computer Science & Information Engineering, Zhejiang Gongshang University, Hangzhou, China. junfu@zjgsu.edu.cn

ABSTRACT
Biologically-inspired models and algorithms are considered as promising sensor array signal processing methods for electronic noses. Feature selection is one of the most important issues for developing robust pattern recognition models in machine learning. This paper describes an investigation into the classification performance of a bionic olfactory model with the increase of the dimensions of input feature vector (outer factor) as well as its parallel channels (inner factor). The principal component analysis technique was applied for feature selection and dimension reduction. Two data sets of three classes of wine derived from different cultivars and five classes of green tea derived from five different provinces of China were used for experiments. In the former case the results showed that the average correct classification rate increased as more principal components were put in to feature vector. In the latter case the results showed that sufficient parallel channels should be reserved in the model to avoid pattern space crowding. We concluded that 6~8 channels of the model with principal component feature vector values of at least 90% cumulative variance is adequate for a classification task of 3~5 pattern classes considering the trade-off between time consumption and classification rate.

Show MeSH

Related in: MedlinePlus

Binary experiments for (a) Pattern reproduce; (b) Pattern retrieval in a 16-channle KIII model with simple Hebbian learning rule.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3376605&req=5

f3-sensors-12-02818: Binary experiments for (a) Pattern reproduce; (b) Pattern retrieval in a 16-channle KIII model with simple Hebbian learning rule.

Mentions: Figure 3 shows pattern reproduce and retrieval of a 16-channle KIII model with simple Hebbian learning rule. Every subplot presents an amplitude modulated pattern (AM) output of each M1 node in the OB layer. Figure 3(a) is pattern output of learned pattern P = [0 1 0 0 0 0 0 1 0 0 0 0 1 0 0 0]. The consistent channel position and amplitude with pattern P demonstrate a memory recall. Figure 3(b) is pattern output of an incomplete version P’ = [0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0]. Compared with pattern P, there is an AM burst in input-absence channel M1(8) (green curve), demonstrating a good pattern recovery property.


Pattern classification using an olfactory model with PCA feature selection in electronic noses: study and application.

Fu J, Huang C, Xing J, Zheng J - Sensors (Basel) (2012)

Binary experiments for (a) Pattern reproduce; (b) Pattern retrieval in a 16-channle KIII model with simple Hebbian learning rule.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3376605&req=5

f3-sensors-12-02818: Binary experiments for (a) Pattern reproduce; (b) Pattern retrieval in a 16-channle KIII model with simple Hebbian learning rule.
Mentions: Figure 3 shows pattern reproduce and retrieval of a 16-channle KIII model with simple Hebbian learning rule. Every subplot presents an amplitude modulated pattern (AM) output of each M1 node in the OB layer. Figure 3(a) is pattern output of learned pattern P = [0 1 0 0 0 0 0 1 0 0 0 0 1 0 0 0]. The consistent channel position and amplitude with pattern P demonstrate a memory recall. Figure 3(b) is pattern output of an incomplete version P’ = [0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0]. Compared with pattern P, there is an AM burst in input-absence channel M1(8) (green curve), demonstrating a good pattern recovery property.

Bottom Line: The principal component analysis technique was applied for feature selection and dimension reduction.In the former case the results showed that the average correct classification rate increased as more principal components were put in to feature vector.We concluded that 6~8 channels of the model with principal component feature vector values of at least 90% cumulative variance is adequate for a classification task of 3~5 pattern classes considering the trade-off between time consumption and classification rate.

View Article: PubMed Central - PubMed

Affiliation: College of Computer Science & Information Engineering, Zhejiang Gongshang University, Hangzhou, China. junfu@zjgsu.edu.cn

ABSTRACT
Biologically-inspired models and algorithms are considered as promising sensor array signal processing methods for electronic noses. Feature selection is one of the most important issues for developing robust pattern recognition models in machine learning. This paper describes an investigation into the classification performance of a bionic olfactory model with the increase of the dimensions of input feature vector (outer factor) as well as its parallel channels (inner factor). The principal component analysis technique was applied for feature selection and dimension reduction. Two data sets of three classes of wine derived from different cultivars and five classes of green tea derived from five different provinces of China were used for experiments. In the former case the results showed that the average correct classification rate increased as more principal components were put in to feature vector. In the latter case the results showed that sufficient parallel channels should be reserved in the model to avoid pattern space crowding. We concluded that 6~8 channels of the model with principal component feature vector values of at least 90% cumulative variance is adequate for a classification task of 3~5 pattern classes considering the trade-off between time consumption and classification rate.

Show MeSH
Related in: MedlinePlus