Local normalized linear summation kernel for fast and robust recognition

Kernel-based methods are effective for object detection and recognition. However, the computational cost when using kernel functions is high, except when using linear kernels. To realize fast and robust recognition, we apply normalized linear kernels to local regions of a recognition target, and the kernel outputs are integrated by summation. This kernel is referred to as a local normalized linear summation kernel. Here, we show that kernel-based methods that employ local normalized linear summation kernels can be computed by a linear kernel of local normalized features. Thus, the computational cost of the kernel is nearly the same as that of a linear kernel and much lower than that of radial basis function (RBF) and polynomial kernels. The effectiveness of the proposed method is evaluated in face detection and recognition problems, and we confirm that our kernel provides higher accuracy with lower computational cost than RBF and polynomial kernels. In addition, our kernel is also robust to partial occlusion and shadows on faces since it is based on the summation of local kernels.     

Greedy versus social: resource-competing oscillator network as a model of amoeba-based neurocomputer

A single-celled amoeboid organism, the true slime mold Physarum polycephalum, exhibits rich spatiotemporal oscillatory behavior and sophisticated computational capabilities. The authors previously created a biocomputer that incorporates the organism as a computing substrate to search for solutions to combinatorial optimization problems. With the assistance of optical feedback to implement a recurrent neural network model, the organism changes its shape by alternately growing and withdrawing its photosensitive branches so that its body area can be maximized and the risk of being illuminated can be minimized. In this way, the organism succeeded in finding the optimal solution to the four-city traveling salesman problem with a high probability. However, it remains unclear how the organism collects, stores, and compares information on light stimuli using the oscillatory dynamics. To study these points, we formulate an ordinary differential equation model of the amoeba-based neurocomputer, considering the organism as a network of oscillators that compete for a fixed amount of intracellular resource. The model, called the “Resource-Competing Oscillator Network (RCON) model,” reproduces well the organism’s experimentally observed behavior, as it generates a number of spatiotemporal oscillation modes by keeping the total sum of the resource constant. Designing the feedback rule properly, the RCON model comes to face a problem of optimizing the allocation of the resource to its nodes. In the problem-solving process, “greedy” nodes having the highest competitiveness are supposed to take more resource out of other nodes. However, the resource allocation pattern attained by the greedy nodes cannot always achieve a “socially optimal” state in terms of the public cost. We prepare four test problems including a tricky one in which the greedy pattern becomes “socially unfavorable” and investigate how the RCON model copes with these problems. Comparing problem-solving performances of the oscillation modes, we show that there exist some modes often attain socially favorable patterns without being trapped in the greedy one.     

Unsupervised WSD by Finding the Predominant Sense Using Context as a Dynamic Thesaurus

We present and analyze an unsupervised method for Word Sense Disambiguation (WSD). Our work is based on the method presented by McCarthy et al. in 2004 for finding the predominant sense of each word in the entire corpus. Their maximization algorithm allows weighted terms (similar words) from a distributional thesaurus to accumulate a score for each ambiguous word sense, i.e., the sense with the highest score is chosen based on votes from a weighted list of terms related to the ambiguous word. This list is obtained using the distributional similarity method proposed by Lin Dekang to obtain a thesaurus. In the method of McCarthy et al., every occurrence of the ambiguous word uses the same thesaurus, regardless of the context where the ambiguous word occurs. Our method accounts for the context of a word when determining the sense of an ambiguous word by building the list of distributed similar words based on the syntactic context of the ambiguous word. We obtain a top precision of 77.54% of accuracy versus 67.10% of the original method tested on SemCor. We also analyze the effect of the number of weighted terms in the tasks of finding the Most Frecuent Sense (MFS) and WSD, and experiment with several corpora for building the Word Space Model.     

Amoeba-based Chaotic Neurocomputing: Combinatorial Optimization by Coupled Biological Oscillators

We demonstrate a neurocomputing system incorporating an amoeboid unicellular organism, the true slime mold Physarum, known to exhibit rich spatiotemporal oscillatory behavior and sophisticated computational capabilities. Introducing optical feedback applied according to a recurrent neural network model, we induce that the amoeba’s photosensitive branches grow or degenerate in a network-patterned chamber in search of an optimal solution to the traveling salesman problem (TSP), where the solution corresponds to the amoeba’s stably relaxed configuration (shape), in which its body area is maximized while the risk of being illuminated is minimized.Our system is capable of reaching the optimal solution of the four-city TSP with a high probability. Moreover, our system can find more than one solution, because the amoeba can coordinate its branches’ oscillatory movements to perform transitional behavior among multiple stable configurations by spontaneously switching between the stabilizing and destabilizing modes. We show that the optimization capability is attributable to the amoeba’s fluctuating oscillatory movements. Applying several surrogate data analyses, we present results suggesting that the amoeba can be characterized as a set of coupled chaotic oscillators.

Parametric polynomial spectral factorization using the sum of roots and its application to a control design problem

This paper presents an algebraic approach to polynomial spectral factorization, an important mathematical tool in signal processing and control. The approach exploits an intriguing relationship between the theory of Gröbner bases and polynomial spectral factorization which can be observed through the sum of roots, and allows us to perform polynomial spectral factorization in the presence of real parameters. It is discussed that parametric polynomial spectral factorization enables us to express quantities such as the optimal cost in terms of parameters and the sum of roots. Furthermore an optimization method over parameters is suggested that makes use of the results from parametric polynomial spectral factorization and also employs two quantifier elimination techniques. This proposed approach is demonstrated in a numerical example of a particular control problem.     

Pose-Invariant Facial Expression Recognition Using Variable-Intensity Templates

In this paper, we propose a method for pose-invariant facial expression recognition from monocular video sequences. The advantage of our method is that, unlike existing methods, our method uses a simple model, called the variable-intensity template, for describing different facial expressions. This makes it possible to prepare a model for each person with very little time and effort. Variable-intensity templates describe how the intensities of multiple points, defined in the vicinity of facial parts, vary with different facial expressions. By using this model in the framework of a particle filter, our method is capable of estimating facial poses and expressions simultaneously. Experiments demonstrate the effectiveness of our method. A recognition rate of over 90% is achieved for all facial orientations, horizontal, vertical, and in-plane, in the range of ±40 degrees, ±20 degrees, and ±40 degrees from the frontal view, respectively. Electronic Supplementary Material  The online version of this article () contains supplementary material, which is available to authorized users.     

A new water-resistant snow index for the detection and mapping of snow cover on a global scale

An up-to-date spatio-temporal change analysis of global snow cover is essential for better understanding of climate–hydrological interactions. The normalized difference snow index (NDSI) is a widely used algorithm for the detection and estimation of snow cover. However, NDSI cannot discriminate between snow cover and water bodies without use of an external water mask. A stand-alone methodology for robust detection and mapping of global snow cover is presented by avoiding external dependency on the water mask. A new spectral index called water-resistant snow index (WSI) with the capability of exhibiting significant contrast between snow cover and other cover types, including water bodies, was developed. WSI uses the normalized difference between the value and hue obtained by transforming red, green, and blue, (RGB) colour composite images comprising red, green, and near-infrared bands into a hue, saturation, and value (HSV) colour model. The superiority of WSI over NDSI is confirmed by case studies conducted in major snow regions globally. Snow cover was mapped by considering monthly variation in snow cover and availability of satellite data at the global scale. A snow cover map for the year 2013 was produced at the global scale by applying the random walker algorithm in the WSI image supported by the reference data collected from permanent snow-covered and non-snow-covered areas. The resultant snow-cover map was compared to snow cover estimated by existing maps: MODIS Land Cover Type Product (MCD12Q1 v5.1, 2012), Global Land Cover by National Mapping Organizations (GLCNMO v2.0, 2008), and European Space Agency’s GlobCover 2009. A significant variation in snow cover as estimated by different maps was noted, and was was attributed to methodological differences rather than annual variation in snow cover. The resultant map was also validated with reference data, with 89.46% overall accuracy obtained. The WSI proposed in the research is expected to be suitable for seasonal and annual change analysis of global snow cover.     

Probabilistic nearest neighbor query processing on distributed uncertain data

A nearest neighbor (NN) query, which returns the most similar object to a user-specified query object, plays an important role in a wide range of applications and hence has received considerable attention. In many such applications, e.g., sensor data collection and location-based services, objects are inherently uncertain. Furthermore, due to the ever increasing generation of massive datasets, the importance of distributed databases, which deal with such data objects, has been growing. One emerging challenge is to efficiently process probabilistic NN queries over distributed uncertain databases. The straightforward approach, that each local site forwards its own database to the central server, is communication-expensive, so we have to minimize communication cost for the NN object retrieval. In this paper, we focus on two important queries, namely top-k probable NN queries and probabilistic star queries, and propose efficient algorithms to process them over distributed uncertain databases. Extensive experiments on both real and synthetic data have demonstrated that our algorithms significantly reduce communication cost.     

Performance of SOR methods on modern vector and scalar processors

The building-cube method (BCM) is a new generation algorithm for CFD simulations. The basic idea of BCM is to simplify the algorithm in all stages of flow computation to achieve large-scale simulations. Calculation of a pressure field using the Successive Over Relaxation (SOR) method consumes most of the total execution time required for BCM. In this paper, effective implementations on modern vector and scalar processors are investigated. NEC SX-9 and Intel Nehalem-EX are the latest vector and scalar processors. Those processors have much higher peak performances than their previous-generation processors. However, their memory bandwidth improvement cannot catch up with the performance improvement of processors. This is the so-called memory wall problem. In our paper, we discuss optimization techniques for implementation of the SOR method based on architectural characteristics of these modern processors, and evaluate their effects on the sustained performances of these processors for BCM.