A Novel Sparsity Reconstruction Method from Poisson Data for 3D Bioluminescence Tomography

In this paper, we consider 3D Bioluminescence tomography (BLT) source reconstruction from Poisson data in three dimensional space. With a priori information of sources sparsity and MAP estimation of Poisson distribution, we study the minimization of Kullback-Leihbler divergence with ℓ ¹ and ℓ ⁰ regularization. We show numerically that although several ℓ ¹ minimization algorithms are efficient for compressive sensing, they fail for BLT reconstruction due to the high coherence of the measurement matrix columns and high nonlinearity of Poisson fitting term. Instead, we propose a novel greedy algorithm for ℓ ⁰ regularization to reconstruct sparse solutions for BLT problem. Numerical experiments on synthetic data obtained by the finite element methods and Monte-Carlo methods show the accuracy and efficiency of the proposed method.     

Stock trading strategy creation using GP on GPU

This paper investigates the speed improvements available when using a graphics processing unit (GPU) for evaluation of individuals in a genetic programming (GP) environment. An existing GP system is modified to enable parallel evaluation of individuals on a GPU device. Several issues related to implementing GP on GPU are discussed, including how to perform tree-based GP on a device without recursion support, as well as the effect that proper memory layout can have on speed increases when using CUDA-enabled nVidia GPU devices. The specific GP implementation is designed to evolve stock trading strategies using technical analysis indicators. The second goal of this research is to investigate the possible improvement in performance when training individuals on a larger number of stocks and training days. This increased training size (nearly 100,000 training points) is enabled due to the speedups realized by GPU evaluation. Several different scenarios were used to test various speed optimizations of GP evaluation on the GPU device, with a peak speedup factor of over 600 (when compared to sequential evaluation on a 2.4 GHz CPU). Also, it is found that increasing the number of stocks and the length of the training period can result in higher out-of-training testing profitability.     

Detection of sudden pedestrian crossings for driving assistance systems

In this paper, we study the problem of detecting sudden pedestrian crossings to assist drivers in avoiding accidents. This application has two major requirements: to detect crossing pedestrians as early as possible just as they enter the view of the car-mounted camera and to maintain a false alarm rate as low as possible for practical purposes. Although many current sliding-window-based approaches using various features and classification algorithms have been proposed for image-/video-based pedestrian detection, their performance in terms of accuracy and processing speed falls far short of practical application requirements. To address this problem, we propose a three-level coarse-to-fine video-based framework that detects partially visible pedestrians just as they enter the camera view, with low false alarm rate and high speed. The framework is tested on a new collection of high-resolution videos captured from a moving vehicle and yields a performance better than that of state-of-the-art pedestrian detection while running at a frame rate of 55 fps.     

Relevance feature mapping for content-based multimedia information retrieval

This paper presents a novel ranking framework for content-based multimedia information retrieval (CBMIR). The framework introduces relevance features and a new ranking scheme. Each relevance feature measures the relevance of an instance with respect to a profile of the targeted multimedia database. We show that the task of CBMIR can be done more effectively using the relevance features than the original features. Furthermore, additional performance gain is achieved by incorporating our new ranking scheme which modifies instance rankings based on the weighted average of relevance feature values. Experiments on image and music databases validate the efficacy and efficiency of the proposed framework.     

Mining the change of customer behavior in fuzzy time-interval sequential patterns

Comprehending changes of customer behavior is an essential problem that must be faced for survival in a fast-changing business environment. Particularly in the management of electronic commerce (EC), many companies have developed on-line shopping stores to serve customers and immediately collect buying logs in databases. This trend has led to the development of data-mining applications. Fuzzy time-interval sequential pattern mining is one type of serviceable data-mining technique that discovers customer behavioral patterns over time. To take a shopping example, (Bread, Short, Milk, Long, Jam), means that Bread is bought before Milk in a Short period, and Jam is bought after Milk in a Long period, where Short and Long are predetermined linguistic terms given by managers. This information shown in this example reveals more general and concise knowledge for managers, allowing them to make quick-response decisions, especially in business. However, no studies, to our knowledge, have yet to address the issue of changes in fuzzy time-interval sequential patterns. The fuzzy time-interval sequential pattern, (Bread, Short, Milk, Long, Jam), became available in last year; however, is not a trend this year, and has been substituted by (Bread, Short, Yogurt, Short, Jam). Without updating this knowledge, managers might map out inappropriate marketing plans for products or services and dated inventory strategies with respect to time-intervals. To deal with this problem, we propose a novel change mining model, MineFuzzChange, to detect the change in fuzzy time-interval sequential patterns. Using a brick-and-mortar transactional dataset collected from a retail chain in Taiwan and a B2C EC dataset, experiments are carried out to evaluate the proposed model. We empirically demonstrate how the model helps managers to understand the changing behaviors of their customers and to formulate timely marketing and inventory strategies.     

Crusta: A new virtual globe for real-time visualization of sub-meter digital topography at planetary scales

Virtual globes are becoming ubiquitous in the visualization of planetary bodies and Earth specifically. While many of the current virtual globes have proven to be quite useful for remote geologic investigation, they were never designed for the purpose of serving as virtual geologic instruments. Their shortcomings have become more obvious as earth scientists struggle to visualize recently released digital elevation models of very high spatial resolution (0.5-1 m²/sample) and extent (>2000 km²). We developed Crusta as an alternative virtual globe that allows users to easily visualize their custom imagery and more importantly their custom topography. Crusta represents the globe as a 30-sided polyhedron to avoid distortion of the display, in particular the singularities at the poles characteristic of other projections. This polyhedron defines 30 “base patches,” each being a four-sided region that can be subdivided to an arbitrarily fine grid on the surface of the globe to accommodate input data of arbitrary resolution, from global (BlueMarble) to local (tripod LiDAR), all in the same visualization. We designed Crusta to be dynamic with the shading of the terrain surface computed on-the-fly when a user manipulates his point-of-view. In a similarly interactive fashion the globe's surface can be exaggerated vertically. The combination of the two effects greatly improves the perception of shape. A convenient pre-processing tool based on the GDAL library facilitates importing a number of data formats into the Crusta-specific multi-scale hierarchies that enable interactive visualization on a range of platforms from laptops to immersive geowalls and caves. The main scientific user community for Crusta is earth scientists, and their needs have been driving the development.