Hyeoncheol Kim

An MLP-based feature subset selection for HIV-1 protease cleavage site analysis

OBJECTIVE In recent years, several machine learning approaches have been applied to modeling the specificity of the human immunodeficiency virus type 1 (HIV-1) protease cleavage domain. However, the high dimensional domain dataset contains a small number of samples, which could misguide classification modeling and its interpretation. Appropriate feature selection can alleviate the problem by eliminating irrelevant and redundant features, and thus improve prediction performance. METHODS We introduce a new feature subset selection method, FS-MLP, that selects relevant features using multi-layered perceptron (MLP) learning. The method includes MLP learning with a training dataset and then feature subset selection using decompositional approach to analyze the trained MLP. Our method is able to select a subset of relevant features in high dimensional, multi-variate and non-linear domains. RESULTS Using five artificial datasets that represent four data types, we verified the FS-MLP performance with seven other feature selection methods. Experimental results showed that the FS-MLP is superior at high dimensional, multi-variate and non-linear domains. In experiments with HIV-1 protease cleavage dataset, the FS-MLP selected a set of 14 highly relevant features among 160 original features. On a validation set of 131 test instances, classifiers that used the 14 features showed about 95% accuracy which outperformed other seven methods in terms of accuracy and the number of features. CONCLUSIONS Our experimental results indicate that the FS-MLP is effective in analyzing multi-variate, non-linear and high dimensional datasets such as HIV-1 protease cleavage dataset. The 14 relevant features which were selected by the FS-MLP provide us with useful insights into the HIV-1 cleavage site domain as well. The FS-MLP is a useful method for computational sequence analysis in general.

Short communication: Specificity rule discovery in HIV-1 protease cleavage site analysis

Several machine learning algorithms have recently been applied to modeling the specificity of HIV-1 protease. The problem is challenging because of the three issues as follows: (1) datasets with high dimensionality and small number of samples could misguide classification modeling and its interpretation; (2) symbolic interpretation is desirable because it provides us insight to the specificity in the form of human-understandable rules, and thus helps us to design effective HIV inhibitors; (3) the interpretation should take into account complexity or dependency between positions in sequences. Therefore, it is necessary to investigate multivariate and feature-selective methods to model the specificity and to extract rules from the model. We have tested extensively various machine learning methods, and we have found that the combination of neural networks and decompositional approach can generate a set of effective rules. By validation to experimental results for the HIV-1 protease, the specificity rules outperform the ones generated by frequency-based, univariate or black-box methods.

A Personalized English vocabulary learning system based on cognitive abilities related to foreign language proficiency

This paper proposes a novel of a personalized Computer Assisted Language Learning (CALL) system based on learner’s cognitive abilities related to foreign language proficiency. In this CALL system, a strategy of retrieval learning, a method of learning memory cycle, and a method of repeated learning are applied for effective vocabulary memorization. The system is designed to offer personalized learning based on cognitive abilities related to the human language process. For this, the proposed CALL system has a cognitive diagnosis module which can measure five types of cognitive abilities. The results of this diagnosis are used to create dynamic learning scenarios for personalized learning and to evaluate user performance in the learning. This system is also designed in order to have users be able to create learning word lists and to share them simply with various functions based on open APIs. Additionally, through experiments, it has shown that this system helps students to learn English vocabulary effectively and enhances their foreign language skills.

Computationally Efficient Heuristics for If-Then Rule Extraction from Freed-Forward Neural Networks

In this paper, we address computational complexity issues of decompositional approaches to if-then rule extraction from feed-forward neural networks. We also introduce a computationally effcient technique based on ordered-attributes. It reduces search space significantly and finds valid and general rules for single nodes in the networks. Empirical results are shown.

A development of learning widget on m-learning and e-learning environments

This article describes the development of learning widget on m-learning and e-learning environments. A widget is a small, simple and useful application supporting user-oriented contents. The user may select and install widgets that are convenient as well as an auto-updating application including weather or calendar. These widgets are especially more useful, because they are able to be installed on a mobile device, a website or a desktop computer. If we take advantage of widgets for education, we may use this learning tool for delivering and pulling learning contents, essences of lessons or word learning. To that end, we developed an effective learning widget and then verified its usability, usefulness and effectiveness for m-learning and e-learning. That is, we evaluated the learning widget with a heuristic evaluation method. We identified 72 interface problems by using a set of 10 usability criteria or heuristics. In addition, we considered how to design the learning widget with consideration given to devices on m-learning and e-learning. Moreover, we experimented by conducting a pilot test with 34 students, a field test with 60 teachers and technology acceptance model (TAM) analysis with 15 teachers. We verified the effectiveness and usefulness of learning with a questionnaire, a quiz and TAM, where the subjects, after using the learning widget in real learning activities, rated the widgets efficacy. The result shows that the learning widget is useful for m-learning and e-learning environments.

Eyes of a wiki: automated navigation map

There are many potential uses of a Wiki within a community-based digital library. Users share individual ideas to build up community knowledge by efficient and effective collaborative authoring and communications that a Wiki provides. In our study, we investigated how the community knowledge is organized into a knowledge structure that users can access and modify efficiently. Since a Wiki provides users with freedom of editing any pages, a Wiki site increases and changes dynamically. We also developed a tool that helps users to navigate easily in the dynamically changing link structure. In our experiment, it is shown that the navigation tool fosters Wiki users to figure out the complex site structure more easily and thus to build up more well-structured community knowledge base. We also show that a Wiki with the navigation tool improves collaborative learning in a web-based e-learning environment.

Real-time video watermarking techniques robust against re-encoding

This paper presents a real-time watermarking codec that is robust against re-encoding attacks for high-definition videos. The codec uses a segmentation function and texture detector techniques for applying real-time watermarking to human visual systems. Experimental results confirm that the proposed scheme satisfies the requirements of invisibility, real-time processing, and robustness against format conversion and low bit-rate encoding. The proposed algorithm has the advantages of simplicity, flexibility, and low computational burden; thus, it is a suitable candidate for many novel and interesting applications such as video fingerprinting for set-top boxes, Internet protocol television, personal video recorders, and satellite boxes.

Predictability of rules in HIV-1 protease cleavage site analysis

Symbolic rules play an important role in HIV-1 protease cleavage site prediction. Recently, some studies have done on extraction of the prediction rules with some success. In this paper, we demonstrated a decompositional approach for rule extraction from nonlinear neural networks. We also compared the prediction rules to the ones extracted by other approaches and methods. Empirical experiments are also shown.

Prediction rule generation of MHC class i binding peptides using ANN and GA

A new method is proposed for generating if-then rules to predict peptide binding to class I MHC proteins, from the amino acid sequence of any protein with known binders and non-binders. In this paper, we present an approach based on artificial neural networks (ANN) and knowledge-based genetic algorithm (KBGA) to predict the binding of peptides to MHC class I molecules. Our method includes rule extraction from a trained neural network and then enhancing the extracted rules by genetic evolution. Experimental results show that the method could generate new rules for MHC class I binding peptides prediction.

Information-based pruning for interesting association rule mining in the item response dataset

Frequency-based mining of association rules sometimes suffers rule quality problems. In this paper, we introduce a new measure called surprisal that estimates the informativeness of transactional instances and attributes. We eliminate noisy and uninformative data using the surprisal first, and then generate association rules of good quality. Experimental results show that the surprisal-based pruning improves quality of association rules in question item response datasets significantly.