Jung-Ying Wang

Prediction and evolutionary information analysis of protein solvent accessibility using multiple linear regression

A multiple linear regression method was applied to predict real values of solvent accessibility from the sequence and evolutionary information. This method allowed us to obtain coefficients of regression and correlation between the occurrence of an amino‐acid residue at a specific target and its sequence neighbor positions on the one hand, and the solvent accessibility of that residue on the other. Our linear regression model based on sequence information and evolutionary models was found to predict residue accessibility with 18.9% and 16.2% mean absolute error respectively, which is better than or comparable to the best available methods. A correlation matrix for several neighbor positions to examine the role of evolutionary information at these positions has been developed and analyzed. As expected, the effective frequency of hydrophobic residues at target positions shows a strong negative correlation with solvent accessibility, whereas the reverse is true for charged and polar residues. The correlation of solvent accessibility with effective frequencies at neighboring positions falls abruptly with distance from target residues. Longer protein chains have been found to be more accurately predicted than their smaller counterparts. Proteins 2005.

SVM-Cabins: prediction of solvent accessibility using accumulation cutoff set and support vector machine.

A number of methods for predicting levels of solvent accessibility or accessible surface area (ASA) of amino acid residues in proteins have been developed. These methods either predict regularly spaced states of relative solvent accessibility or an analogue real value indicating relative solvent accessibility. While discrete states of exposure can be easily obtained by post prediction assignment of thresholds to the predicted or computed real values of ASA, the reverse, that is, obtaining a real value from quantized states of predicted ASA, is not straightforward as a two‐state prediction in such cases would give a large real valued errors. However, prediction of ASA into larger number of ASA states and then finding a corresponding scheme for real value prediction may be helpful in integrating the two approaches of ASA prediction. We report a novel method of obtaining numerical real values of solvent accessibility, using accumulation cutoff set and support vector machine. This so‐called SVM‐Cabins method first predicts discrete states of ASA of amino acid residues from their evolutionary profile and then maps the predicted states onto a real valued linear space by simple algebraic methods. Resulting performance of such a rigorous approach using 13‐state ASA prediction is at least comparable with the best methods of ASA prediction reported so far. The mean absolute error in this method reaches the best performance of 15.1% on the tested data set of 502 proteins with a coefficient of correlation equal to 0.66. Since, the method starts with the prediction of discrete states of ASA and leads to real value predictions, performance of prediction in binary states and real values are simultaneously optimized. Proteins 2007.

Model selection of SVMs using GA approach

A new automatic search methodology for model selection of support vector machines, based on the GA-based tuning algorithm, is proposed to search for the adequate hyperparameters of SVMs. In our method, each chromosome indicates a group of hyperparameters, and the population is a collection of chromosomes. Experimental results show that our method performs superiorly on time cost, performance and stability. Our algorithm requires only the evaluation of an objective function to guide its search with no additional derivative or auxiliary knowledge required. In addition, the encoding of chromosomes makes the implementation of multiple hyperparameters tuning simpler.

Recognition of Human Actions Using Motion Capture Data and Support Vector Machine

This paper presents a human action recognition system based on motion capture features and support vector machine (SVM). We use 43 optical markers distributing on body and extremities to track the movement of human actions. In our system 21 different types of action are recognized. Applying SVM for the recognition of human action the overall prediction accuracy achieves to 84.1% when using the three-fold cross validation on the training set. Another purpose of this study is to find out which skeleton points are important for human action recognition. The experimental results show that the skeleton points of head, hands and feet are the most important features for recognition of human actions.

Game AI: Simulating Car Racing Game by Applying Pathfinding Algorithms

In this paper, two modified A* algorithms to effectively solve the pathfinding problem in a static obstacles racing game are proposed. Three real speedways of Formula one (F1) are selected as our game speedways, to simulate and analyze our study. The first modified A* algorithm uses a line-of-sight algorithm to reduce the waypoints found by the original A* algorithm; about 97% waypoints in the speedways of F1 in Turkey, Italy and Hungary could be removed. The second modified A* algorithm improves the performance of original A* algorithm by heuristically considering the truth that the game-controlled car should steer itself towards. That is to say, we could reduce the lap times by only checking three waypoints in front of the car, instead of checking four waypoints (up, down, left and right) in the original A* algorithm. Finally, a more general dynamic pathfinding algorithm which can solve the random obstacles avoidance problem in a racing game is also proposed.

An effective method of pathfinding in a car racing game

The most common artificial intelligence in a racing game is waypoint navigation by carefully placing points (nodes) in the game environment to move the game-controlled characters between each point. This kind of pathfinding algorithm is a very time consuming and CPU intensive problem. In this paper, we present an algorithm to reduce the consuming time and simplify the implementation process. We simply put two collision detection points in front of the right and left side of the car, and then utilize the trick of collision detection by different color to automatic move the cars around the racetrack. In our algorithm, two adjusted variables have been carefully studied. The first variable is the collision detection distance, and the second variable is the turning radian of the car when collision is detected. The experimental results show that the lap time decreases as the collision detection distance increases, when the car do not stick by any curve. That is, the lap time decreases while the collision detection point is far from the car. It is because the car has more time to perceive the edge of motordrome to respond the collision. Similarly, the turning radian of the car increases, the time needs for turning the curve decreases, causing a decrease in lap time. However, when the cars change the racetrack from straight to the more curvature curve, or meet the successive curve in a short distance, large value of the collision detection distance or turning radian will cause the car stick in the corner of racetrack.

Effects of Online Synchronous Instruction with an Attention Monitoring and Alarm Mechanism on Sustained Attention and Learning Performance.

ABSTRACT Many studies have shown that learners’ sustained attention strongly affects e-learning performance, particularly during online synchronous instruction. This work thus develops a novel attention monitoring and alarm mechanism (AMAM) based on brainwave signals to improve learning performance via monitoring the attention state of individual learners and helping online instructors or teaching assistants to improve the sustained attention levels of learners with low-attention states as they perform online synchronous instruction activities. Totally, 83 and 65 Grade 7 students were randomly assigned to the experimental and control groups that respectively underwent online synchronous instruction with and without AMAM support. Analytical results reveal that the experimental group of learners exhibited significantly better learning performance and sustained attention than those in the control group, verifying that the AMAM efficiently promotes the learning performance and sustained attention of learners. Moreover, the proposed AMAM was more helpful in improving the learning performance of female learners than those of male learners and improved the sustained attention of both male and female learners. Furthermore, the sustained attention, frequency of attention alarms, and learning performance of the learners in the experimental group were strongly correlated, and the sustained attention and frequency of attention alarms strongly predicted learning performance.

Designing Brain Training Games and Evaluating the Usability between Young and Elderly

There is few researches focus on investigating game players experience and preference between the young and elderly. In this study, four different braining training games designed by us are used to investigate the effects of age and game preference. In total, 136 participants were randomly selected, and they are separated into two age groups-young (N=90) and old (N=46). The experimental results demonstrate differential age effects in braining training games performance. Our results show that no matter what types of games young adults get better performance than the older adults and they all reach a significant difference. Meanwhile, due to the interface design may affect the willingness of older adults to play the game. Therefore, the usability is also studied by the design of games control keys in a car racing game. The experimental results show that, older people respond relatively more slowly than the young people. Therefore, simpler interface design of brain training game is more suitable for older adults. In addition, the survey results indicated that the fruit catching game is loved by most of old people (reached 41.2%). The main reasons are without problem-solving stress and easy to play-player only need to control the right and left arrow keys. In other words, no psychological burden and simple interface design are the key factors in design popular games for elderly. However, young group is more like car racing game than others (28.6%). Since it is a playable game and they get a feeling of satisfaction when they reach a lower lap time. Finally, the brain age reference diagrams for the four developed games are proposed. Players can use these diagrams to understand their capability with respect to young and old group.

Comparison of game experience and preferences between young and elderly

With an increasingly aging population, currently, Taiwan has become an aging society. Because of cognitive decline in the elderly, many brain training games target this population. But there is few researches focus on investigating game players experience and preference of the elderly. In this study, our primary purpose is to develop and design brain training games as an intermediary to explore the differences of response speed and memory between elderly and youth. In this study, three brain training games are designed and developed by us. There are a 2D shooting game, a color sequence game and a card matching game. Two age group (young and old) and three types of game (training for response speed, training for memory and training for hybrid of both) constituting a 2 × 3 mixed factorial design are investigated. In total, 125 participants were randomly selected, and they are separated into two age groups-young (N=86) and old (N=39). Experimental results show that no matter what types of games-training for speed, training for memory or training for hybrid of both, young people get better performance than the aged and they all reach a significant difference. The survey results indicated that 2D shooting game is loved by most of old people (reached 82%). The main reason is easy to play-player only need to control the up, down, right, left and fire keys, and has no psychological burden. In other words, no memory pressure and no time limit pressure are the key factors in design popular games for elderly. However, young group is more like the color sequence game than others (41.9%). Since the playing type and design of color sequence game is different with most of the normal games. Finally, the brain age reference diagrams for the three developed games are proposed. Players can use these diagrams to understand their capability with respect to young and old group.

Dynamic difficulty adjustment by fuzzy rules using in a neural network controlled game

This paper describes a series of experiments using the offline trained artificial neural networks (ANN). The ANN acts as an embedded game agent in a shooting game to control the nonplayer character (NPC). The training datasets of ANN are constructed by three different levels of players (expert, medium and beginner players). And then the three different levels training datasets are used to train three different levels ANN, respectively. Meanwhile, the optimal neurons of the hidden layer and the suitable period of training time is obtained by the method of three fold cross validation. In addition, a comparison between ANN and two traditional game AI - finite state machine (FSM) and computer random controlled method, is also implemented in this study. The simulated results show that ANN can get better winning rate than FSM and random method. Meanwhile, ANN obtains a pretty good human-like simulation results. Finally, a fuzzy rules-based approach is utilized to do the dynamic game difficulty adjustment. The experimental results show that the adaptive mechanism developed in this study could dynamic balance the equilibrium of game difficulty. All these, enhance the replayability of the game.