Search

Summary: SMotif is a server that identifies important structural segments or motifs for a given protein structure(s) based on conservation of both sequential as well as important structural features such as solvent inaccessibility, secondary structural content, hydrogen bonding pattern and residue packing. This server also provides three-dimensional orientation patterns of the identified motifs in terms of inter-motif distances and torsion angles. These motifs may form the common core and therefore, can also be employed to design and rationalize protein engineering and folding experiments. Availability: SMotif server is available via the URL http://caps.ncbs.res.in/SMotif/index.html. Contact: chakraba@mail.nih.gov, mini@ncbs.res.in or EPNSugan@ntu.edu.sg Supplementary information: Supplementary data are available at Bioinformatics online.

The purpose of this paper is to demonstrate how a structural matching approach can be used to perfonn effective rotational invariant fingerprint identification. In this approach, each of the exiracted features is correlated with Live of its nearest neighbouring features to form a local feature gmup for a first-stage matching. After that, the feature with the highest match is used as a central feature whereby all the other features are correlated to form a global feature group for a second.stage matching. The correlation between the features is in terms of distance and relative angle. This approach actually make the matching method rotational invariant A substantial amount of testing was carried out and it shows that this matching technique is capable of matching the four basic fingerprint patterns with an average matching time of4 seconds on a 66Mhz, 486 DX personal computer.

In the past, the Hopfield network has been employed to solve pattern recognition problems by subgraph isomorphism which naturally constrains the scene to have at most one occurrence of any object model. Recently, the author proposed a novel programming procedure to generate a homomorphic mapping which enables simultaneous recognition of multiple instances of any particular object model in the scene (P.N. Suganthan, 1995; 1995). However, in order to generate the desired homomorphic mapping, a number of parameters have to be fine tuned. A self-organising Hopfield network is introduced that learns the constraint parameters using a Liapunov indirect method based learning approach.

Background Odorant binding proteins (OBPs) are believed to shuttle odorants from the environment to the underlying odorant receptors, for which they could potentially serve as odorant presenters. Although several sequence based search methods have been exploited for protein family prediction, less effort has been devoted to the prediction of OBPs from sequence data and this area is more challenging due to poor sequence identity between these proteins. Results In this paper, we propose a new algorithm that uses Regularized Least Squares Classifier (RLSC) in conjunction with multiple physicochemical properties of amino acids to predict odorant-binding proteins. The algorithm was applied to the dataset derived from Pfam and GenDiS database and we obtained overall prediction accuracy of 97.7% (94.5% and 98.4% for positive and negative classes respectively). Conclusion Our study suggests that RLSC is potentially useful for predicting the odorant binding proteins from sequence-derived properties irrespective of sequence similarity. Our method predicts 92.8% of 56 odorant binding proteins non-homologous to any protein in the swissprot database and 97.1% of the 414 independent dataset proteins, suggesting the usefulness of RLSC method for facilitating the prediction of odorant binding proteins from sequence information.

A method of using particle swarm optimisation (PSO) algorithms to optimise the element spacing and lengths of Yagi-Uda antennas is presented. SuperNEC, an object-oriented version of the numerical electromagnetic code (NEC-2) is used to evaluate the performance of various Yagi-Uda antenna designs. In order to show the capabilities of the PSO algorithm in Yagi-Uda antenna design, three different antenna design cases are optimised for various performance specifications. The three objectives considered are gain only, gain and input impedance only, and gain, input impedance and relative sidelobe level (rSLL). To alleviate the premature convergence problem of PSO, a novel learning strategy is employed. Each design problem is optimised using three variants of PSO algorithms, namely the modified PSO, fitness-distance ratio PSO (FDR-PSO), and comprehensive learning PSO (CLPSO). For the purpose of comparison and benchmarking, equally spaced arrays, genetic algorithm optimised antenna design, and computational intelligence optimised antenna design are considered. The results clearly show that the CLPSO is a robust and useful optimisation tool for designing Yagi antennas for the desired target specifications.

3-dimensional domain swapping is a mechanism where two or more protein molecules form higher order oligomers by exchanging identical or similar subunits. Recently, this phenomenon has received much attention in the context of prions and neurodegenerative diseases, due to its role in the functional regulation, formation of higher oligomers, protein misfolding, aggregation etc. While 3-dimensional domain swap mechanism can be detected from three-dimensional structures, it remains a formidable challenge to derive common sequence or structural patterns from proteins involved in swapping. We have developed a SVM-based classifier to predict domain swapping events using a set of features derived from sequence and structural data. The SVM classifier was trained on features derived from 150 proteins reported to be involved in 3D domain swapping and 150 proteins not known to be involved in swapped conformation or related to proteins involved in swapping phenomenon. The testing was performed using 63 proteins from the positive dataset and 63 proteins from the negative dataset. We obtained 76.33% accuracy from training and 73.81% accuracy from testing. Due to high diversity in the sequence, structure and functions of proteins involved in domain swapping, availability of such an algorithm to predict swapping events from sequence and structure-derived features will be an initial step towards identification of more putative proteins that may be involved in swapping or proteins involved in deposition disease. Further, the top features emerging in our feature selection method may be analysed further to understand their roles in the mechanism of domain swapping.

This work describes a method of designing a reconfigurable dual-beam antenna array using a novel concurrent PSO (CONPSO) and different variants of FDR-PSO algorithms. The design problem is to find element excitations that will result in a sector pattern main beam with low side lobes with additional requirement that the same excitation amplitudes applied to the array with zero-phase should result in a high directivity, low side lobe and pencil shaped main beam. In order to test the performances of various PSO algorithms, the results of real-coded genetic algorithm is considered, for the purpose of comparison. Experiment results clearly indicate the better performance of various FDR-PSO and proposed CONPSO algorithms over simple PSO and genetic algorithms in terms of solution quality, and consistency. The proposed CONPSO algorithm is very much suitable for implementation in parallel computer.

Knowledge of three dimensional structure is essential to understand the function of a protein. Although the overall fold is made from the whole details of its sequence, a small group of residues, often called as structural motifs, play a crucial role in determining the protein fold and its stability. Identification of such structural motifs requires sufficient number of sequence and structural homologs to define conservation and evolutionary information. Unfortunately, there are many structures in the protein structure databases have no homologous structures or sequences. In this work, we report an SVM method, SMpred, to identify structural motifs from single protein structure without using sequence and structural homologs. SMpred method was trained and tested using 132 proteins domains containing 581 motifs. SMpred method achieved 78.79% accuracy with 79.06% sensitivity and 78.53% specificity. The performance of SMpred was evaluated with MegaMotifBase using 188 proteins containing 1161 motifs. Out of 1161 motifs, SMpred correctly identified 1503 structural motifs reported in MegaMotifBase. Further, we showed that SMpred is useful approach for the length deviant superfamilies and single member superfamilies. This result suggests the usefulness of our approach for facilitating the identification of structural motifs in protein structure in the absence of sequence and structural homologs. The dataset and executable for the SMpred algorithm is available at http://www3.ntu.edu.sg/home/EPNSugan/index_files/SMpred.htm.

Evolutionary algorithms are successfully developed to handle the challenges in solving optimization problems with complex landscapes. Differential evolution proves its efficiency as a powerful evolutionary algorithm to solve complex optimization problems with diverse characteristics. In this paper, we aim at designing an enhanced evolutionary algorithm that embeds Differential Evolution in Stochastic Fractal Search. Stochastic Fractal Search is developed recently as a powerful metaheuristic algorithm that imitates the natural phenomenon of growth and uses the diffusion process based on random fractals. In this paper, we introduce a new adjustment to the Diffusion Process of Stochastic Fractal Search. The proposed algorithm namely, SFS-DPDE-GW, uses Differential Evolution in the Diffusion Process along with the Gaussian Walks to enhance the search. To validate the performance of our algorithm, a challenging test suite of 30 benchmark functions from the IEEE CEC2014 real parameter single objective competition is used. The proposed combination clearly enhances the performance of Stochastic Fractal Search and increases the efficiency of the update process which was incorporated after Diffusion process. Comparative studies show that the new algorithm has a superior performance compared to the original Stochastic Fractal Search and other recent state-of-the-art algorithms.

In pattern classification area, linear discriminant analysis (LDA) is one of the most traditional methods to find a linear solution to the feature extraction problem, which maximise the ratio between between-class scatter and the within class scatter (Fisher's criterion). We propose a variant of LDA which incorporates the class conjunctions thereby making LDA more robust for the problems in which the within class scatter is quite different from one class to another, while retaining all the merits of conventional LDA. We also integrate an evolutionary search procedure in our algorithm to make it more unbiased to the training samples and to improve the robustness.

Feature selection is an important preprocessing technique for many pattern recognition problems. When the number of features is very large while the number of samples is relatively small as in the micro-array data analysis, feature selection is even more important. This paper proposes a novel feature selection method to perform gene selection from DNA microarray data. The method originates from the least squares support vector machine (LSSVM). The particle swarm optimization (PSO) algorithm is also employed to perform optimization. Experimental results clearly demonstrate good and stable performance of the proposed method.

Many real world problems can be formulated as optimization problems with various parameters to be optimized. Some problems only have one objective to be optimized, some may have multiple objectives to be optimized at the same time and some need to be optimized subjecting to one or more constraints. Thus numerous optimization algorithms have been proposed to solve these problems. Particle Swarm Optimizer (PSO) is a relatively new optimization algorithm which has shown its strength in the optimization world. This thesis presents two PSO variants, Comprehensive Learning PSO and Dynamic Multi-Swarm PSO, which have good global search ability and can solve complex multi-modal problems for single objective optimization. The latter one' is extended to solve constrained optimization and multi-objective optimization problems successfully with a novel constraint-handling mechanism and a novel updating criterion respectively. Subsequently, the Dynamic Multi-Swarm PSO is applied to determine the Bragg wavelengths of the sensors in an FBG sensor network and a tree search structure is designed to improve the accuracy and reduce the computation cost. DOCTOR OF PHILOSOPHY (EEE)

With the development of economy and the change of people's investment consciousness, stock investment has been an important part in daily life, and stock forecasting has also been the focus of investors and financial researchers. Since the income and risk of stock investment is directly proportional, how to establish a relatively high speed and high accurate stock market forecasting model is significant and also has practical value for financial investors. My research mainly introduces the development of time series forecasting, and also gives a review of existing algorithms. In this dissertation, I mainly focus on four forecasting methods namely ANN, SVM, RVFL and RF, along with the experiments we have set up for evaluation. Moreover, a method for signal decomposition called empirical mode decomposition is used to improve the accuracy, along with five datasets that are utilized to test and verify the effectiveness of the proposed method. Master of Science (Computer Control and Automation)

Actually, exchange rate is a kind of important data in economy. There is immense economic information which can be found through exchange rate. Actually, many researchers have used several methods and done experiments to forecast the exchange rate. Machine learning is popular and can be used in many fields. A common use of it is doing some forecasting works in time-series because it can be used to deal with regression problems. Neural Network is an efficient machine learning method to solve forecasting problems. Random Vector Functional Link Network, at the same time, also shows good performance for solving the forecasting problems. In addition, Random Forest and Support Vector Regression also perform well. In this paper, we use machine learning methods mentioned above to forecast the exchange rate between five different currencies and the United States dollar. Keywords: Exchange Rate, Time Series Forecasting, Machine Learning, Neural Networks, Random Forest, Support Vector Regression, Random Vector Functional Link Network Master of Science (Computer Control and Automation)

Many optimization functions have complex landscapes with multiple global or local optima. In order to solve such problems, niching evolutionary algorithms were introduced. The “niching” concept in evolutionary algorithms was brought from the ecological “niches”. It describes the roles that different individuals take when there are several optima to pursue. Niching gives growth to diverse promising species in the population, making it possible to locate multiple optima in a multimodal landscape. In this thesis, a literature review on evolutionary algorithms and several classes of niching methods is presented. After that, a simulation-based comparative study is carried out using hybrid composition test functions with multiple global optima. Three popular niching techniques with binary genetic algorithms are examined for their searching ability, accuracy and computation speed in solving the hybrid composition problems. The number of functions evaluations is employed as the main performance measure. It has been observed that the performance of the niching methods varies with problems, while methods that belong to the same class have shared characteristics. MASTER OF ENGINEERING (EEE)

Evolutionary Algorithms (EAs) are population based algorithms that can tackle complex optimization problems with minimal information about the characteristics of the problem. The performance of Evolutionary Programming (EP), a veteran of the evolutionary computation community depends mostly on the mutation operation, where an offspring is produced from the parent by adding a scaled random number distribution. In EP, the scale factor is referred to as the strategy parameter and is self-adapted using a lognormal adaptation. The abrupt reduction in the strategy parameter values due to the lognormal self-adaptation may result in the premature convergence of the search process. To overcome the drawbacks of lognormal self-adaptation, we propose an adaptive EP (AEP). AEP is different from EP in terms of initialization and adaptation of the strategy parameter values. The parameters are initialized scaled to the search range and are adapted based on the search performance in the previous few generations. Doctor of Philosophy

57 p. The dissertation analyses the procedure for training the SVM for an imbalance multiclass dataset and two-class dataset and thereby maximize the prediction accuracy. One against all approach is followed for the multiclass problem and standard binary SVM for the two-class dataset. Experiments were performed to find the prediction accuracy using the proposed algorithm. The algorithm is tested on five datasets having 5261 samples (1444 features), 5261 samples (61 features), 768 samples, 197 samples (23 features), 267 samples (45 features). The probability estimates and also the decision functions values are found for the multiclass datasets. The one against all accuracies and prediction accuracies for the datasets considered are tabulated. The classification accuracy using the proposed method is tabulated below for each dataset. The SCOP datasets were classified with accuracy of 55.638 and 55.42 percentages. The SCOP datasets are multiclass datasets, whereas the two-class datasets as Pima dataset, Parkinson's dataset and SPECTF heart dataset were classified with the accuracy of 77, 93.2254 and 80.769 percentage respectively. The Algorithm used in this project needs to be tested on more datasets in the future. Master of Science (Computer Control and Automation)

Unprecedented amount of data coming from various high-throughput techniques in biomedical research has presented challenges to computational intelligence (CI) methods to develop efficient algorithms for extraction of knowledge from them. These challenges can be put into one or more of the following different CI perspectives: Feature Selection or feature extraction. Prediction (Classification/Regression). Discovery of new classes (Clustering). This thesis has studied problems in computational biology domain with above perspectives and developed efficient CI methods for them. The research work can be categorized into three parts. First part discusses multi-class classification problems. A framework of class-wise feature selection with One-versus-all-support vector machine (OVA-SVM) classifier is studied. In this framework three methods of probability conversion methods of decision function values of SVM and their ensemble are studied and compared. The proposed framework for multiclass classification has two distinguishing characteristics. First, it uses class-wise optimized features and second, decisions of different SVMs classifiers are coupled with probability estimates to make final prediction. The proposed approach is used for multi-class classification of several biological datasets. It obtained better or at least competitive predictive accuracy to the current best methods for the selected datasets. In all three cases, brief analysis is performed on class-wise selected features to verify their biological relevance to the associated class. Doctor of Philosophy

As palm-print recognition systems gain popularity in the market, there is a growing need to design efficient and accurate algorithms. There is a possibility especially in big organizations that large number of people will make use of the system. It is required that the system is robust enough to handle such a large amount of data. The number of users could be of the order of several hundreds or even thousands. In this work, a 3 stage recognition process has been developed which ensures fast operation while maintaining a high level of accuracy and is capable of handling large databases. The two distinct levels of matching are coarse-level matching and fine-level matching. The first 2 stages are coarse level matching stages which make use of hand geometry and standard deviation values of the local intensity levels. The 3rd stage is a fine level matching stage which uses Gabor filtering to generate Palm-print Phase and Orientation Code (PPOC) and matches two feature sets using modified Hamming distance. The results obtained are very impressive and clearly show the advantage of using a 3-stage process. Master of Science (Computer Control and Automation)

Evolutionary programming (EP) has been applied to solve many numerical global optimization problems successfully during the past four decades. The ability of the algorithm to seach for global optimum depends on the mutation strategy adapted. During the last four decades several mutation operators such as Gaussian and Cauchy (a special case of Levy’s mutation) have been used with EP. According to no free lunch theorem, it is impossible for EP with a single mutation operator to outperform on every problem. For example, Classical EP (CEP) that uses Gaussian mutation is better at search in a local neighborhood while Fast EP (FEP) that uses Cauchy mutation is very good in a large neighborhood. This may give an impression that CEP is better in unimodal problems while FEP is better on multimodal problems/ but in a particular problem due to lack of proper knowledge wbout the global optimum point we need to have global exploration along with local exploiting during every generation of the evolution process. Motivated by these observations, we propose and Ensemble of mutation operator (Gaussian and Cauchy) to solve CEC 2005 problems by using evolutionary programming (EP) algorithm. This eliminates the need to perform the trail-and-error search for the best mutation operator, but also enable us to benefit from the match between the exploration-exploitation stages of the search process. The distinguished feature of Ensemble algorithm is the effective usage of every function call. The offspring population produced by a particular mutation operator may dominate the others at particular stage of the optimization process. Furthermore, an offspring produced by a particular mutation operator may be rejected by its own population, but could be accepted by the population of other mutation method. We employ two different mutation operators (Gaussian and Cauchy) present in the literature to form the ensemble. Master of Science (Computer Control and Automation)

65 p. Portfolio optimization is a multi-objective, non-linear optimization problem for maximum return with minimum risk. It normally has a huge number of input variables (assets) and numerous local optima; Therefore mathematical derivatives based optimization is very difficult or impossible to be applied. Genetic algorithms (GA) have been used for this kind of problem, as it is good to deal with optimization involving a large number of inputs and non-linear multi-model objectives. But performance of GA optimization degrades when the number of inputs of the problem increases. Master of Science (Computer Control and Automation)