Seam carving is a content-aware image processing approach that has been successfully applied to resizing or removing objects from digital images. Seam-carved images are hard to identify from the original image, making this detection method an important and attractive research topic. Existing methods are based on steganography attacks or statistical features, and their trained and tested models are all constructed using support vector machines (SVMs). The trained models of these methods take a long time to build from a limited number of images. This paper presents a new patch-based sobel operator (PSO) method using SVM and extreme learning machines (ELMs) based on the patch analysis method with square-based features. The ELM adopts five types of neurons to obtain five prediction results, which are then combined using the jury voting scheme in order to improve the accuracy. In addition, combinatorial fusion is first used to choose two of five prediction results according to the diversity rank/score graph, and these two results are then combined to make the final decision. The PSO method in the ELM with jury voting and combinatorial fusion achieves accuracies of 70.60-98.80% and 95.58-99.36% for 10-50% seam-carved and seam-insertion images, respectively. Additionally, for the PSO method, the ELM required only 1.2% of the training time required by the SVM. In conclusion, the PSO method in the ELM is useful for constructing trained models from a large number of images.

High-end graphics processing units (GPUs), such as NVIDIA Fermi/Tesla series cards, are widely applied to the high performance computing fields in a decade. NVIDIA releases Tegra K1, called Jetson TK1, which contains 4 ARM Cortex-A15 CPUs and 192 CUDA cores (Kepler GPU) is an embedded board with low cost, low power consumption, and high applicability advantages for several specific applications. In the previous work, we have constructed a bioinformatics platform based on single NVIDIA Jetson TK1, and ClustalWtk tool for multiple sequence alignments was designed on this platform. With more and more biological sequences generated and high computing power requirement, it is necessary to construct a cluster platform based on multiple NVIDIA Jetson TK1. In this paper, a bioinformatics platform based on a cluster of multiple NVIDIA Jetson TK1 is constructed. Besides, two job assignment modes are proposed in this platform: one is a “user selected” mode and another is an “automatic assigned” mode. ClustalWtk tool also is ported on this platform with these two modes, respectively.

Mobile computing devices facilitate the completion of tasks in all aspects of our daily lives. However, mobile displays have been recognized to have adverse impacts on our vision, and they may even cause severe injures to our eyes. Mobile devices cause visual discomfort, not only because of screen glare but also as a result of handheld shake and vibration. To address the aforementioned concerns, we assessed the implementation of an anti-shake mechanism for mobile display to provide users with a comfortable interaction experience and to minimize negative side effects of screen vibrations. The overall process of implementing this mechanism includes analyzing the accelerometer data, predicting the magnitude of device displacement, and adjusting the position of the display area to compensate for vibrations. In this paper, we propose a fuzzy inference model for motion prediction and display stabilization. We verified this model with tumbling E eye exams and subjective 10-point scale questionnaires. The experimental results reveal that the anti-shake mechanism successfully decreases sight fatigue when applied to mobile devices used in dynamically changing environments.

A major difficulty in a search-based problem-solving process is the task of searching the potentially huge search space resulting from the exponential growth of states. State explosion rapidly occupies memory and increases computation time. Although various heuristic search algorithms have been developed to solve problems in a reasonable time, there is no efficient method to construct heuristic functions. In this work, we propose a method by which a neural network can be iteratively trained to form an efficient heuristic function. An adaptive heuristic search procedure is involved in the training iterations. This procedure reduces the evaluation values of the states that are involved in the currently known best solution paths. By doing so, the promising states are continuously moved forward. The adapted heuristic values are fed back to neural networks; thus, a well-trained network function can find the near-best solutions quickly. To demonstrate this method, we solved the fifteen-puzzle problem. Experimental results showed that the solutions obtained by our method were very close to the shortest path, and both the number of explored nodes and the search time were significantly reduced.

Algorithm visualization is a unique research topic that integrates engineering skills such as computer graphics, system programming, database management, computer networks, etc., to facilitate algorithmic researchers in testing their ideas, demonstrating new findings, and teaching algorithm design in the classroom. Within the broad applications of algorithm visualization, there still remain performance issues that deserve further research, e.g., system portability, collaboration capability, and animation effect in 3D environments. Using modern technologies of Java programming, we develop an algorithm visualization and debugging system, dubbed GeoBuilder, for geometric computing. The GeoBuilder system features Java's promising portability, engagement of collaboration in algorithm development, and automatic camera positioning for tracking 3D geometric objects. In this paper, we describe the design of the GeoBuilder system and demonstrate its applications.

Geometric algorithm visualization techniques are very important to algorithmic research in geometric computing. The relevant applications include geometric algorithm development, testing, demonstration and teaching. Within these topics, there still remain performance issues to improve system portability, animation effect and so on. In this paper, we present a geometric algorithm visualization system that is featured by Java's portability and the "dynamic decision of camera position" for 3D geometric algorithm visualization.

Benchmark tests are often used to evaluate the quality of products by a set of common criteria. In this paper we describe a computational problem solving environment based on open source codes and an algorithm benchmark system, which is embedded in the environment as a plug-in system. The algorithm benchmark system can be used to compare the performance of various algorithms or to evaluate the behavior of an algorithm with different input instances. The current implementation allows users to compare or evaluate algorithms written in C/C++. Some examples of the algorithm benchmark system that evaluates the memory utilization, time complexity and the output of algorithms are also presented. Algorithm benchmark impresses the learning effect; students can not only comprehend the performance of respective algorithms but also write their own programs to challenge the best known results.

The traveling salesman problem is difficult to solve by traditional genetic algorithms because of the requirement that each node must be visited exactly once. In response to this critical requirement, many researchers used specialized operators to adapt traditional genetic algorithms. Although some of these operators are useful, they are ad hoc. We propose a priority-based encoding scheme instead. We assign priorities to all the edges and then perform a greedy algorithm to find a suboptimal solution. The greedy algorithm constructs a legal tour and the priority encoding makes it possible to follow traditional genetic evolution. This approach retains generality in applications and also gains remarkable experimental results.

Hysteresis is a memory effect frequently observed in physical research. The output of a hysteresis system is independent of input speed. This property, known as rate independence, significantly distinguishes hysteresis from short-term memory effects. This work conducts a numerical simulation to demonstrate that conventional models with short-term memories cannot properly simulate hysteresis trajectories. Subsequently, a novel model is developed to contribute to the field of system modeling. Experimental results confirm that the proposed model can model hysteresis behavior precisely.

Hysteresis is a unique type of dynamic, which contains an important property, rate-independent memory. In addition to other memory-related studies such as time delay neural networks, recurrent networks, and reinforcement learning, rate-independent memory deserves further attention owing to its potential applications. In this paper, we attempt to define hysteretic memory (rate independent memory) and examine whether or not it could be modeled in neural networks. Our analysis results demonstrate that other memory-related mechanisms are not hysteresis systems. A novel neural cell, referred to herein as the propulsive neural unit, is then proposed. The proposed cell is based on a notion related the submemory pool, which accumulates the stimulus and ultimately assists neural networks to achieve model hysteresis. In addition to training by backpropagation, a combination of such cells can simulate given hysteresis trajectories.