Search results for: vector variance.

Authors: H. B. Kekre, Tanuja K. Sarode

Abstract:

Mostly transforms are used for speech data compressions which are lossy algorithms. Such algorithms are tolerable for speech data compression since the loss in quality is not perceived by the human ear. However the vector quantization (VQ) has a potential to give more data compression maintaining the same quality. In this paper we propose speech data compression algorithm using vector quantization technique. We have used VQ algorithms LBG, KPE and FCG. The results table shows computational complexity of these three algorithms. Here we have introduced a new performance parameter Average Fractional Change in Speech Sample (AFCSS). Our FCG algorithm gives far better performance considering mean absolute error, AFCSS and complexity as compared to others.

Keywords: Vector Quantization, Data Compression, Encoding, , Speech coding.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2351

Authors: Driss Guerchi, Fatiha Djebbar

Abstract:

In this work we introduce an efficient method to limit the impact of the hiding process on the quality of the cover speech. Vector quantization of the speech spectral information reduces drastically the number of the secret speech parameters to be embedded in the cover signal. Compared to scalar hiding, vector quantization hiding technique provides a stego signal that is indistinguishable from the cover speech. The objective and subjective performance measures reveal that the current hiding technique attracts no suspicion about the presence of the secret message in the stego speech, while being able to recover an intelligible copy of the secret message at the receiver side.

Keywords: Speech steganography, LSF vector quantization, fast Fourier transform

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1464

Authors: Stephen Kirkup

Abstract:

This paper discusses the implementation of the boundary element method (BEM) on an Excel spreadsheet and how it can be used in teaching vector calculus and simulation. There are two separate spreadheets, within which Laplace equation is solved by the BEM in two dimensions (LIBEM2) and axisymmetric three dimensions (LBEMA). The main algorithms are implemented in the associated programming language within Excel, Visual Basic for Applications (VBA). The BEM only requires a boundary mesh and hence it is a relatively accessible method. The BEM in the open spreadsheet environment is demonstrated as being useful as an aid to teaching and learning. The application of the BEM implemented on a spreadsheet for educational purposes in introductory vector calculus and simulation is explored. The development of assignment work is discussed, and sample results from student work are given. The spreadsheets were found to be useful tools in developing the students’ understanding of vector calculus and in simulating heat conduction.

Keywords: Boundary element method, laplace equation, vector calculus, simulation, education.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 929

Authors: S. Esakkirajan, T. Veerakumar, V. Senthil Murugan, R. Sudhakar

Abstract:

This paper presents a new fingerprint coding technique based on contourlet transform and multistage vector quantization. Wavelets have shown their ability in representing natural images that contain smooth areas separated with edges. However, wavelets cannot efficiently take advantage of the fact that the edges usually found in fingerprints are smooth curves. This issue is addressed by directional transforms, known as contourlets, which have the property of preserving edges. The contourlet transform is a new extension to the wavelet transform in two dimensions using nonseparable and directional filter banks. The computation and storage requirements are the major difficulty in implementing a vector quantizer. In the full-search algorithm, the computation and storage complexity is an exponential function of the number of bits used in quantizing each frame of spectral information. The storage requirement in multistage vector quantization is less when compared to full search vector quantization. The coefficients of contourlet transform are quantized by multistage vector quantization. The quantized coefficients are encoded by Huffman coding. The results obtained are tabulated and compared with the existing wavelet based ones.

Keywords: Contourlet Transform, Directional Filter bank, Laplacian Pyramid, Multistage Vector Quantization

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1954

Authors: Aldan A. Sapargaliyev, Yerbol A. Sapargaliyev

Abstract:

Since the invention, the electric machine (EM) can be defined as oEM – one-vector electric machine, as it works due to one-vector inductive coupling with use of one-vector electromagnet. The disadvantages of oEM are large size and limited efficiency at low and medium power applications. This paper describes multi-vector electric machine (mEM) based on multi-vector inductive coupling, which is characterized by the increased surface area of ​​the inductive coupling per EM volume, with a reduced share of inefficient and energy-consuming part of the winding, in comparison with oEM’s. Particularly, it is considered, calculated and compared the performance of three different electrical motors and their power at the same volumes and rotor frequencies. It is also presented the result of calculation of correlation between power density and volume for oEM and mEM. The method of multi-vector inductive coupling enables mEM to possess 1.5-4.0 greater density of power per volume and significantly higher efficiency, in comparison with today’s oEM, especially in low and medium power applications. mEM has distinct advantages, when used in transport vehicles such as electric cars and aircrafts.

Keywords: Electric machine, electric motor, electromagnet, efficiency of electric motor.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 982

Authors: Ahmed Abdolkhalig, Rastko Zivanovic

Abstract:

This paper evaluates the Total Vector Error of an estimated Phasor as define in IEEE C37.118 standard within different medium access in Local Area Networks (LAN). Three different LAN models (CSMA/CD, CSMA/AMP and Switched Ethernet) are evaluated. The Total Vector Error of the estimated Phasor has been evaluated for the effect of Nodes Number under the standardized network Band-width values defined in IEC 61850-9-2 communication standard (i.e. 0.1, 1 and 10 Gbps).

Keywords: Phasor, Local Area Network, Total Vector Error, IEEE C37.118, IEC 61850.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 4096

Authors: M. Debyeche, J.P Haton, A. Houacine

Abstract:

The paper presents a complete discrete statistical framework, based on a novel vector quantization (VQ) front-end process. This new VQ approach performs an optimal distribution of VQ codebook components on HMM states. This technique that we named the distributed vector quantization (DVQ) of hidden Markov models, succeeds in unifying acoustic micro-structure and phonetic macro-structure, when the estimation of HMM parameters is performed. The DVQ technique is implemented through two variants. The first variant uses the K-means algorithm (K-means- DVQ) to optimize the VQ, while the second variant exploits the benefits of the classification behavior of neural networks (NN-DVQ) for the same purpose. The proposed variants are compared with the HMM-based baseline system by experiments of specific Arabic consonants recognition. The results show that the distributed vector quantization technique increase the performance of the discrete HMM system.

Keywords: Hidden Markov Model, Vector Quantization, Neural Network, Speech Recognition, Arabic Language

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2005

Authors: R. Simutis, V. Galvanauskas, D. Levisauskas, J. Repsyte

Abstract:

This paper proposes a bioprocess optimization procedure based on Relevance Vector Regression models and evolutionary programming technique. Relevance Vector Regression scheme allows developing a compact and stable data-based process model avoiding time-consuming modeling expenses. The model building and process optimization procedure could be done in a half-automated way and repeated after every new cultivation run. The proposed technique was tested in a simulated mammalian cell cultivation process. The obtained results are promising and could be attractive for optimization of industrial bioprocesses.

Keywords: Bioprocess optimization, Evolutionary programming, Relevance Vector Regression.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2136

Authors: Wei Zhang, Su-Yan Tang, Yi-Fan Zhu, Wei-Ping Wang

Abstract:

Support vector regression (SVR) has been regarded as a state-of-the-art method for approximation and regression. The importance of kernel function, which is so-called admissible support vector kernel (SV kernel) in SVR, has motivated many studies on its composition. The Gaussian kernel (RBF) is regarded as a “best" choice of SV kernel used by non-expert in SVR, whereas there is no evidence, except for its superior performance on some practical applications, to prove the statement. Its well-known that reproducing kernel (R.K) is also a SV kernel which possesses many important properties, e.g. positive definiteness, reproducing property and composing complex R.K by simpler ones. However, there are a limited number of R.Ks with explicit forms and consequently few quantitative comparison studies in practice. In this paper, two R.Ks, i.e. SV kernels, composed by the sum and product of a translation invariant kernel in a Sobolev space are proposed. An exploratory study on the performance of SVR based general R.K is presented through a systematic comparison to that of RBF using multiple criteria and synthetic problems. The results show that the R.K is an equivalent or even better SV kernel than RBF for the problems with more input variables (more than 5, especially more than 10) and higher nonlinearity.

Keywords: admissible support vector kernel, reproducing kernel, reproducing kernel Hilbert space, support vector regression.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1534

Authors: Daniel I. Morariu, Lucian N. Vintan, Volker Tresp

Abstract:

Text categorization is the problem of classifying text documents into a set of predefined classes. After a preprocessing step, the documents are typically represented as large sparse vectors. When training classifiers on large collections of documents, both the time and memory restrictions can be quite prohibitive. This justifies the application of feature selection methods to reduce the dimensionality of the document-representation vector. In this paper, we present three feature selection methods: Information Gain, Support Vector Machine feature selection called (SVM_FS) and Genetic Algorithm with SVM (called GA_SVM). We show that the best results were obtained with GA_SVM method for a relatively small dimension of the feature vector.

Keywords: Feature Selection, Learning with Kernels, Support Vector Machine, Genetic Algorithm, and Classification.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1655

Authors: Eric Jiang

Abstract:

As a popular rank-reduced vector space approach, Latent Semantic Indexing (LSI) has been used in information retrieval and other applications. In this paper, an LSI-based content vector model for text classification is presented, which constructs multiple augmented category LSI spaces and classifies text by their content. The model integrates the class discriminative information from the training data and is equipped with several pertinent feature selection and text classification algorithms. The proposed classifier has been applied to email classification and its experiments on a benchmark spam testing corpus (PU1) have shown that the approach represents a competitive alternative to other email classifiers based on the well-known SVM and naïve Bayes algorithms.

Keywords: Feature Selection, Latent Semantic Indexing, Text Classification, Vector Space Model.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1838

Authors: Karchung, S. Ruangsinchaiwanich

Abstract:

This paper presents open-loop vector control method of induction motor with space vector pulse width modulation (SVPWM) technique. Normally, the closed loop speed control is preferred and is believed to be more accurate. However, it requires a position sensor to track the rotor position which is not desirable to use it for certain workspace applications. This paper exhibits the performance of three-phase induction motor with the simplest control algorithm without the use of a position sensor nor an estimation block to estimate rotor position for sensorless control. The motor stator currents are measured and are transformed to synchronously rotating (d-q-axis) frame by use of Clarke and Park transformation. The actual control happens in this frame where the measured currents are compared with the reference currents. The error signal is fed to a conventional PI controller, and the corrected d-q voltage is generated. The controller outputs are transformed back to three phase voltages and are fed to SVPWM block which generates PWM signal for the voltage source inverter. The open loop vector control model along with SVPWM algorithm is modeled in MATLAB/Simulink software and is experimented and validated in TMS320F28335 DSP board.

Keywords: Electric drive, induction motor, open-loop vector control, space vector pulse width modulation technique.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 881

Authors: Zeinab Ghasemi, S. Amirhassan Monadjemi, Abbas Vafaei

Abstract:

This paper presents a comparative analysis of a new unsupervised PCA-based technique for steel plates texture segmentation towards defect detection. The proposed scheme called Variance Based Component Analysis or VBCA employs PCA for feature extraction, applies a feature reduction algorithm based on variance of eigenpictures and classifies the pixels as defective and normal. While the classic PCA uses a clusterer like Kmeans for pixel clustering, VBCA employs thresholding and some post processing operations to label pixels as defective and normal. The experimental results show that proposed algorithm called VBCA is 12.46% more accurate and 78.85% faster than the classic PCA.

Keywords: Principal Component Analysis; Variance Based Component Analysis; Defect Detection; Texture Segmentation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1918

Authors: Tapan Trivedi, Pramod Agarwal, Rajendrasinh Jadeja, Pragnesh Bhatt

Abstract:

Space Vector Pulse Width Modulation is popular for variable frequency drives. The method has several advantages over carried based PWM and is computation intensive. The implementation of SVPWM for multilevel inverter requires special attention and at the same time consumes considerable resources. Due to faster processing power and reduced over all computational burden, FPGAs are being investigated as an alternative for other controllers. In this paper, a space vector PWM algorithm is implemented using FPGA which requires less computational area and is modular in structure. The algorithm is verified experimentally for Neutral Point Clamped inverter using FPGA development board xc3s5000-4fg900.

Keywords: Modular structure, Multilevel inverter, Space Vector PWM, Switching States.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2378

Authors: J. S. Kim, N. P. Hong

Abstract:

In this paper, a CMOS vector modulator designed for wireless backhaul system based on 802.11ac is presented. A poly phase filter and sign select switches yield two orthogonal signal paths. Two variable gain amplifiers with strongly reduced phase shift of only ±5 ° are used to weight these paths. It has a phase control range of 360 ° and a gain range of -10 dB to 10 dB. The current drawn from a 1.2 V supply amounts 20.4 mA. Using a 0.13 mm technology, the chip die area amounts 1.47x0.75 mm².

Keywords: CMOS, vector modulator, backhaul, 802.11ac.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2210

Authors: Willem Waegeman, Luc Boullart

Abstract:

Instead of traditional (nominal) classification we investigate the subject of ordinal classification or ranking. An enhanced method based on an ensemble of Support Vector Machines (SVM-s) is proposed. Each binary classifier is trained with specific weights for each object in the training data set. Experiments on benchmark datasets and synthetic data indicate that the performance of our approach is comparable to state of the art kernel methods for ordinal regression. The ensemble method, which is straightforward to implement, provides a very good sensitivity-specificity trade-off for the highest and lowest rank.

Keywords: Ordinal regression, support vector machines, ensemblelearning.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1567

Authors: Jin-Woong Park, Sun-Kyung Kang, Sung-Tae Jung

Abstract:

In this study, we propose a tongue diagnosis method which detects the tongue from face image and divides the tongue area into six areas, and finally generates tongue coating ratio of each area. To detect the tongue area from face image, we use ASM as one of the active shape models. Detected tongue area is divided into six areas widely used in the Korean traditional medicine and the distribution of tongue coating of the six areas is examined by SVM(Support Vector Machine). For SVM, we use a 3-dimensional vector calculated by PCA(Principal Component Analysis) from a 12-dimentional vector consisting of RGB, HIS, Lab, and Luv. As a result, we detected the tongue area stably using ASM and found that PCA and SVM helped raise the ratio of tongue coating detection.

Keywords: Active Shape Model, Principal Component Analysis, Support Vector Machine, Tongue diagnosis

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1824

Authors: J. Ruiz-Pinales, J. J. Acosta-Reyes, A. Salazar-Garibay, R. Jaime-Rivas

Abstract:

In this paper, we present a new method for incorporating global shift invariance in support vector machines. Unlike other approaches which incorporate a feature extraction stage, we first scale the image and then classify it by using the modified support vector machines classifier. Shift invariance is achieved by replacing dot products between patterns used by the SVM classifier with the maximum cross-correlation value between them. Unlike the normal approach, in which the patterns are treated as vectors, in our approach the patterns are treated as matrices (or images). Crosscorrelation is computed by using computationally efficient techniques such as the fast Fourier transform. The method has been tested on the ORL face database. The tests indicate that this method can improve the recognition rate of an SVM classifier.

Keywords: Face recognition, support vector machines, shiftinvariance, image registration.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1688

Authors: Süha Yılmaz, Melih Turgut

Abstract:

The curves, of which the square of the distance between the two points equal to zero, are called minimal or isotropic curves [4]. In this work, first, necessary and sufficient conditions to be a Pseudo Helix, which is a special case of such curves, are presented. Thereafter, it is proven that an isotropic curve-s position vector and pseudo curvature satisfy a vector differential equation of fourth order. Additionally, In view of solution of mentioned equation, position vector of pseudo helices is obtained.

Keywords: Classical Differential Geometry, Euclidean space, Minimal Curves, Isotropic Curves, Pseudo Helix.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1920

Authors: K. Freedman

Abstract:

In the present study, a support vector machine (SVM) learning approach to character recognition is proposed. Simple feature detectors, similar to those found in the human visual system, were used in the SVM classifier. Alphabetic characters were rotated to 8 different angles and using the proposed cognitive model, all characters were recognized with 100% accuracy and specificity. These same results were found in psychiatric studies of human character recognition.

Keywords: Character recognition, cognitive model, support vector machine learning.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1836

Authors: Tatjana Eitrich, Bruno Lang

Abstract:

This work deals with aspects of support vector learning for large-scale data mining tasks. Based on a decomposition algorithm that can be run in serial and parallel mode we introduce a data transformation that allows for the usage of an expensive generalized kernel without additional costs. In order to speed up the decomposition algorithm we analyze the problem of working set selection for large data sets and analyze the influence of the working set sizes onto the scalability of the parallel decomposition scheme. Our modifications and settings lead to improvement of support vector learning performance and thus allow using extensive parameter search methods to optimize classification accuracy.

Keywords: Support Vector Machines, Shared Memory Parallel Computing, Large Data

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1535

Authors: Hadi Sadoghi Yazdi, Tahereh Royani, Mehri Sadoghi Yazdi, Sohrab Effati

Abstract:

In this paper, a new version of support vector regression (SVR) is presented namely Fuzzy Cost SVR (FCSVR). Individual property of the FCSVR is operation over fuzzy data whereas fuzzy cost (fuzzy margin and fuzzy penalty) are maximized. This idea admits to have uncertainty in the penalty and margin terms jointly. Robustness against noise is shown in the experimental results as a property of the proposed method and superiority relative conventional SVR.

Keywords: Support vector regression, Fuzzy input, Fuzzy cost.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1324

Authors: Laila Y. Fannas, Ahmed Y. Ben Sasi

Abstract:

This research presents a handwritten signature recognition based on angle feature vector using Artificial Neural Network (ANN). Each signature image will be represented by an Angle vector. The feature vector will constitute the input to the ANN. The collection of signature images will be divided into two sets. One set will be used for training the ANN in a supervised fashion. The other set which is never seen by the ANN will be used for testing. After training, the ANN will be tested for recognition of the signature. When the signature is classified correctly, it is considered correct recognition otherwise it is a failure.

Keywords: Signature Recognition, Artificial Neural Network, Angle Features.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2448

Authors: Mihaela Popescu, Alexandru Bitoleanu, Mircea Dobriceanu

Abstract:

This paper deals with the current space-vector decomposition in three-phase, three-wire systems on the basis of some case studies. We propose four components of the current spacevector in terms of DC and AC components of the instantaneous active and reactive powers. The term of supplementary useless current vector is also pointed out. The analysis shows that the current decomposition which respects the definition of the instantaneous apparent power vector is useful for compensation reasons only if the supply voltages are sinusoidal. A modified definition of the components of the current is proposed for the operation under nonsinusoidal voltage conditions.

Keywords: Active current, Active filtering, p–q theory, Reactive current.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1467

Authors: Kostas Metaxiotis

Abstract:

Portfolio optimization is one of the most important topics in finance. This paper proposes a mean–variance–skewness (MVS) portfolio optimization model. Traditionally, the portfolio optimization problem is solved by using the mean–variance (MV) framework. In this study, we formulate the proposed model as a three-objective optimization problem, where the portfolio's expected return and skewness are maximized whereas the portfolio risk is minimized. For solving the proposed three-objective portfolio optimization model we apply an adapted version of the non-dominated sorting genetic algorithm (NSGAII). Finally, we use a real dataset from FTSE-100 for validating the proposed model.

Keywords: Evolutionary algorithms, portfolio optimization, skewness, stock selection.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1313

Authors: Chan Weng Howe, Mohd Saberi Mohamad

Abstract:

Protein residue contact map is a compact representation of secondary structure of protein. Due to the information hold in the contact map, attentions from researchers in related field were drawn and plenty of works have been done throughout the past decade. Artificial intelligence approaches have been widely adapted in related works such as neural networks, genetic programming, and Hidden Markov model as well as support vector machine. However, the performance of the prediction was not generalized which probably depends on the data used to train and generate the prediction model. This situation shown the importance of the features or information used in affecting the prediction performance. In this research, support vector machine was used to predict protein residue contact map on different combination of features in order to show and analyze the effectiveness of the features.

Keywords: contact map, protein residue contact, support vector machine, protein structure prediction

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1838

Authors: Hani M. Aburas

Abstract:

“Dengue" is an African word meaning “bone breaking" because it causes severe joint and muscle pain that feels like bones are breaking. It is an infectious disease mainly transmitted by female mosquito, Aedes aegypti, and causes four serotypes of dengue viruses. In recent years, a dramatic increase in the dengue fever confirmed cases around the equator-s belt has been reported. Several conventional indices have been designed so far to monitor the transmitting vector populations known as House Index (HI), Container Index (CI), Breteau Index (BI). However, none of them describes the adult mosquito population size which is important to direct and guide comprehensive control strategy operations since number of infected people has a direct relationship with the vector density. Therefore, it is crucial to know the population size of the transmitting vector in order to design a suitable and effective control program. In this context, a study is carried out to report a new statistical index, ABURAS Index, using Poisson distribution based on the collection of vector population in Jeddah Governorate, Saudi Arabia.

Keywords: Poisson distribution, statistical index, prediction, Aedes aegypti.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1871

Authors: Miljan B. Petrović, Dušan B. Petrović, Goran S. Nikolić

Abstract:

This paper describes a method for AWGN (Additive White Gaussian Noise) variance estimation in noisy stochastic signals, referred to as Multiplicative-Noising Variance Estimation (MNVE). The aim was to develop an estimation algorithm with minimal number of assumptions on the original signal structure. The provided MATLAB simulation and results analysis of the method applied on speech signals showed more accuracy than standardized AR (autoregressive) modeling noise estimation technique. In addition, great performance was observed on very low signal-to-noise ratios, which in general represents the worst case scenario for signal denoising methods. High execution time appears to be the only disadvantage of MNVE. After close examination of all the observed features of the proposed algorithm, it was concluded it is worth of exploring and that with some further adjustments and improvements can be enviably powerful.

Keywords: Noise, signal-to-noise ratio, stochastic signals, variance estimation.

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 2207

Authors: Tatjana Eitrich, Bruno Lang

Abstract:

This work deals with aspects of support vector machine learning for large-scale data mining tasks. Based on a decomposition algorithm for support vector machine training that can be run in serial as well as shared memory parallel mode we introduce a transformation of the training data that allows for the usage of an expensive generalized kernel without additional costs. We present experiments for the Gaussian kernel, but usage of other kernel functions is possible, too. In order to further speed up the decomposition algorithm we analyze the critical problem of working set selection for large training data sets. In addition, we analyze the influence of the working set sizes onto the scalability of the parallel decomposition scheme. Our tests and conclusions led to several modifications of the algorithm and the improvement of overall support vector machine learning performance. Our method allows for using extensive parameter search methods to optimize classification accuracy.

Keywords: Support Vector Machine Training, Multi-ParameterKernels, Shared Memory Parallel Computing, Large Data

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1391

Authors: Wei Zhang, Yao-Yu Li, Yi-Fan Zhu, Qun Li, Wei-Ping Wang

Abstract:

It-s known that incorporating prior knowledge into support vector regression (SVR) can help to improve the approximation performance. Most of researches are concerned with the incorporation of knowledge in the form of numerical relationships. Little work, however, has been done to incorporate the prior knowledge on the structural relationships among the variables (referred as to Structural Prior Knowledge, SPK). This paper explores the incorporation of SPK in SVR by constructing appropriate admissible support vector kernel (SV kernel) based on the properties of reproducing kernel (R.K). Three-levels specifications of SPK are studied with the corresponding sub-levels of prior knowledge that can be considered for the method. These include Hierarchical SPK (HSPK), Interactional SPK (ISPK) consisting of independence, global and local interaction, Functional SPK (FSPK) composed of exterior-FSPK and interior-FSPK. A convenient tool for describing the SPK, namely Description Matrix of SPK is introduced. Subsequently, a new SVR, namely Motivated Support Vector Regression (MSVR) whose structure is motivated in part by SPK, is proposed. Synthetic examples show that it is possible to incorporate a wide variety of SPK and helpful to improve the approximation performance in complex cases. The benefits of MSVR are finally shown on a real-life military application, Air-toground battle simulation, which shows great potential for MSVR to the complex military applications.

Keywords: admissible support vector kernel, reproducing kernel, structural prior knowledge, motivated support vector regression

Procedia APA BibTeX Chicago EndNote Harvard JSON MLA RIS XML ISO 690 PDF Downloads 1575