A new generalized LVQ algorithm via harmonic to minimum distance measure transition

Qin, A. K.; Suganthan, Ponnuthurai Nagaratnam; Liang, Jianxing

doi:10.1109/icsmc.2004.1401294

Cited by 3 publications

(5 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…( 3) are replaced by harmonic average distances [74]. For more details the reader is referred to [54].…”

Section: Lvq Methods Based On Margin Maximizationmentioning

confidence: 99%

“…Other LVQ improvement deals with the initialization sensitivity of the original LVQ algorithms and GLVQ [21,31,53,54]. Recent extensions of the LVQ family of algorithms substitute the Euclidean distance with more general metric structures such as: weighted Euclidean metrics [18], adaptive relevance matrix metrics [61], pseudo-Euclidean metrics [22], and similarity measures in kernel feature space that lead to kernelized versions of LVQ [52].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A review of learning vector quantization classifiers

Nova

Estévez

2013

Neural Comput & Applic

111

117

View full text Add to dashboard Cite

show abstract

“…( 3) are replaced by harmonic average distances [74]. For more details the reader is referred to [54].…”

Section: Lvq Methods Based On Margin Maximizationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A review of learning vector quantization classifiers

Nova

Estévez

2013

Neural Comput & Applic

111

117

View full text Add to dashboard Cite

show abstract

“…We have proposed a new algorithm called MLVQ1 (Modified Learning Vector Quantization), which is based on the classic LVQ algorithm (Qin, Suganthan, & Liang, 2004) and which is used during the training step (Sayoud, 2003).…”

Section: New Prosodic Methodsmentioning

confidence: 99%

“…LVQ1 is one of the LVQ algorithms (Qin, Suganthan, & Liang, 2004) used in classification. So, assume that a number of 'codebook vectors' m i (free parameter vectors) are placed into the input space to approximate various domains of the input vector x by their quantized values.…”

Section: Note Thatmentioning

confidence: 99%

Pertinent Prosodic Features for Speaker Identification by Voice

Sayoud

Ouamour

2010

International Journal of Mobile Computing and Multimedia Communications

View full text Add to dashboard Cite

Most existing systems of speaker recognition use “state of the art” acoustic features. However, many times one can only recognize a speaker by his or her prosodic features, especially by the accent. For this reason, the authors investigate some pertinent prosodic features that can be associated with other classic acoustic features, in order to improve the recognition accuracy. The authors have developed a new prosodic model using a modified LVQ (Learning Vector Quantization) algorithm, which is called MLVQ (Modified LVQ). This model is composed of three reduced prosodic features: the mean of the pitch, original duration, and low-frequency energy. Since these features are heterogeneous, a new optimized metric has been proposed that is called Optimized Distance for Heterogeneous Features (ODHEF). Tests of speaker identification are done on Arabic corpus because the NIST evaluations showed that speaker verification scores depend on the spoken language and that some of the worst scores were got for the Arabic language. Experimental results show good performances of the new prosodic approach.

show abstract

“…The NN decision rule assigns to an unclassied sample point the class of the nearest of a set of previously classied points [162]. The prototype based learning algorithm provides a simple and intuitive model while promising generalization performance in pattern classication tasks [163].…”

Section: Nearest Neighbor (Nn)mentioning

confidence: 99%

Data-driven product family design for additive manufacturing

Lei¹

View full text Add to dashboard Cite

Platform based product family design is a promising approach to meet diverse Customer Needs (CNs) and achieve organizational objectives. We have dedicated this work to improve product family design by incorporating advanced information and new manufacturing technologies. Our eort resulted in a data-driven product family design for Additive Manufacturing (AM) method. The proposed data-driven approach used Data Mining (DM) to extract meaningful information from market data. The extracted information was interpreted by advanced machine learning algorithms to form a Decision Support System (DSS), that helps designers make informed decisions for market segmentation and product positioning. Based on the identied market segments, an AM process model for product family design was developed to oer aordable customization for each targeted market segment. Finally, a Utility-Based Compromise Decision Support Problem (u-cDSP) was formulated to serve as a mathematical framework for modeling a multi-objective product family design problem. The thesis highlights data-driven decision making and the opportunities for AM based product family design to operate in a much broader design space that is free from constraints which arise in traditional product family designs from nding a compromise between commonality and product performances. The data-driven product family design for AM method was tested and veried through three distinct case studies. The rst case study focused on the design of a DSS for market segmentation and product positioning based on US automotive market data. The VII proposed DSS automates market segmentation and product positioning and provides a framework for the construction of a robust DSS. In the second case study, we used the proposed method to design a product family of cantilever beams. We found that our process model reects the ability of AM to produce arbitrarily complex structures with virtually no tooling eort, and it makes these powerful properties available to practitioners working in the eld of product family design. The nal case study centered on the design of a dialysis nger pump family. The proposed method translates the benets of AM into improved customization and cost reduction without compromising individual product performances. We created new knowledge in the product family design area by describing the theoretical and empirical validation process of the data-driven product family design for AM method. The main result of this research is a systematic framework which seamlessly integrates AM technologies into product family design to facilitate improved customization. The primary contribution of the framework is a data-driven DSS that advances market segmentation and product positioning. It is expected that the proposed method will redene how we think about customization in product family design. VIII

show abstract

A new generalized LVQ algorithm via harmonic to minimum distance measure transition

Cited by 3 publications

References 6 publications

A review of learning vector quantization classifiers

A review of learning vector quantization classifiers

Pertinent Prosodic Features for Speaker Identification by Voice

Data-driven product family design for additive manufacturing

Contact Info

Product

Resources

About