Kitsuchart Pasupa

Power system stabilizer tuning based on multiobjective design using hierarchical and parallel micro genetic algorithm

In order to achieve the optimal design based on some specific criteria by applying conventional techniques, sequence of design, selected locations of PSSs are critical involved factors. This paper presents a method to simultaneously tune PSSs in multimachine power system using hierarchical genetic algorithm (HGA) and parallel micro genetic algorithm (parallel micro-GA) based on multiobjective function comprising the damping ratio, damping factor and number of PSSs. First, the problem of selecting proper PSS parameters is converted to a simple multiobjective optimization problem. Then, the problem is solved by a parallel micro GA based on HGA. The stabilizers are tuned to simultaneously shift the lightly damped and undamped oscillation modes to a specific stable zone in the s-plane and to self identify the appropriate choice of PSS locations by using eigenvalue-based multiobjective function. Many scenarios with different operating conditions have been included in the process of simultaneous tuning so as to guarantee the robustness and their performance. A 68-bus and 16-generator power system has been employed to validate the effectiveness of the proposed tuning method.

Virtual Screening Using Binary Kernel Discrimination: Effect of Noisy Training Data and the Optimization of Performance

Binary kernel discrimination (BKD) uses a training set of compounds, for which structural and qualitative activity data are available, to produce a model that can then be applied to the structures of other compounds in order to predict their likely activity. Experiments with the MDL Drug Data Report database show that the optimal value of the smoothing parameter, and hence the predictive power of BKD, is crucially dependent on the number of false positives in the training set. It is also shown that the best results for BKD are achieved using one particular optimization method for the determination of the smoothing parameter that lies at the heart of the method and using the Jaccard/Tanimoto coefficient in the kernel function that is used to compute the similarity between a test set molecule and the members of the training set.

Thai sentiment terms construction using the Hourglass of Emotions

Most of the current sentiment analysis techniques classifies emotions into two classes which are positive and negative. Some works classify them as positive, negative and objective (neutral). In fact, there are many kinds of emotions in human mind. Recently, psychological viewpoints have influenced most of the works in sentiment analysis. This psychology perspective was adopted to classify human emotions into a wider range, and in a more accurate manner. This paper reviews the adopted computational representation of emotions the so-called Hourglass of Emotion. This paper also proposes a construction of Thai sentiment resource based on such representation for Thai sentiment term tagging. A preliminary sentiment text tagging result shows that the resource as an ontology can be successfully used to tag sentiment text in Thai children stories.

A simple iterative algorithm for parsimonious binary kernel Fisher discrimination

By applying recent results in optimisation theory variously known as optimisation transfer or majorise/minimise algorithms, an algorithm for binary, kernel, Fisher discriminant analysis is introduced that makes use of a non-smooth penalty on the coefficients to provide a parsimonious solution. The problem is converted into a smooth optimisation that can be solved iteratively with no greater overhead than iteratively re-weighted least-squares. The result is simple, easily programmed and is shown to perform, in terms of both accuracy and parsimony, as well as or better than a number of leading machine learning algorithms on two well-studied and substantial benchmarks.

Exploration-exploitation of eye movement enriched multiple feature spaces for content-based image retrieval

In content-based image retrieval (CBIR) with relevance feedback we would like to retrieve relevant images based on their content features and the feedback given by users. In this paper we view CBIR as an Exploration-Exploitation problem and apply a kernel version of the LinRel algorithm to solve it. By using multiple feature extraction methods and utilising the feedback given by users, we adopt a strategy of multiple kernel learning to find a relevant feature space for the kernel LinRel algorithm. We call this algorithm LinRelMKL. Furthermore, when we have access to eye movement data of users viewing images we can enrich our (multiple) feature spaces by using a tensor kernel SVM. When learning in this enriched space we show that we can significantly improve the search results over the LinRel and LinRelMKL algorithms. Our results suggest that the use of exploration-exploitation with multiple feature spaces is an efficient way of constructing CBIR systems, and that when eye movement features are available, they should be used to help improve CBIR.

An approach to face shape classification for hairstyle recommendation

It is important to choose a good hairstyle for women because it can enhance their beauty, personality, and confidence. One of the most important factors to consider for choosing the right hairstyle is the individuals face shape. An effective face shape classification can be used for constructing a hairstyle recommendation system. This paper presents a classification approach that divides face shapes into 5 different shapes: round, oval, oblong, square, and heart. This approach, which is based on an Active Appearance Model (AAM) and a face segmentation technique, produces a set of features that can be evaluated by several popular machine learning methods, namely, Linear Discriminant Analysis (LDA), Artificial Neural Networks (ANN), and Support Vector Machine (SVM). Our results show that the Support Vector Machine with Radial Basis function kernel was the best algorithm that predicted accurately up to 72%.

Drug screening with Elastic-net multiple kernel learning

We apply Elastic-net Multiple Kernel Learning (MKL) to the MDL Drug Data Report (MDDR) database for the problem of drug screening. We show that combining a set of kernels constructed from fingerprint descriptors, can significantly improve the accuracy of prediction, against a Support Vector Machine trained on each kernel separately. To the best of our knowledge, this is the first application of MKL to the MDDR database for drug screening.

A coefficient comparison of weighted similarity extreme learning machine for drug screening

Machine learning techniques are becoming popular in drug discovery process. It can be used to predict the biological activities of compounds. This paper focuses on virtual screening task. We proposed the Weighted Similarity Extreme Learning Machine algorithm (WELM). It is based on Single Layer Feedforward Neural Network. The algorithm is powerful, iteratively free, and easy to program. In this work, we compared the performance of 17 different types of coefficients with WELM on a well-known dataset in the area of virtual screening named Maximum Unbiased Validation dataset. Moreover, the WELM with different types of coefficients were also compared with the conventional technique-similarity searching. WELM together with Jaccard/Tanimoto were able to achieve the best results on average in most of the activity classes.

Parsimonious Kernel Fisher Discrimination

By applying recent results in optimization transfer, a new algorithm for kernel Fisher Discriminant Analysis is provided that makes use of a non-smooth penalty on the coefficients to provide a parsimonious solution. The algorithm is simple, easily programmed and is shown to perform as well as or better than a number of leading machine learning algorithms on a substantial benchmark. It is then applied to a set of extreme small-sample-size problems in virtual screening where it is found to be less accurate than a currently leading approach but is still comparable in a number of cases.

Hypothesis testing based on observation from Thai sentiment classification

This work focuses on error analyzes from the Support Vector Machine (SVM) classification on Thai children stories at a sentence level. The construction of the Sentiment Term Tagging System (STTS) program allows the researchers to make observations and hypothesize around the areas where most anomalies occur. Three hypotheses, based on terms sentiment chosen for SVM predictions, are evidently proved to hold. In addition, a number of ways to improve the Thai sentiment classification research are suggested, including considerations to add negation into the process, add weighing scheme for different part-of-speech, disambiguate word senses, and update the Thai sentiment resource.