In this paper, we apply a mutation operation based on a multivariate Cauchy distribution to fast evolutionary programming and analyze its effect in terms of various function optimizations. The conventional fast evolutionary programming in-cooperates the univariate Cauchy mutation in order to overcome the slow convergence rate of the canonical Gaussian mutation. For a mutation of n variables, while the conventional method utilizes n independent random variables from a univariate Cauchy distribution, the proposed method adopts n mutually dependent random variables that satisfy a multivariate Cauchy distribution. The multivariate Cauchy distribution naturally has higher probabilities of generating random variables in inter-variable regions than the univariate Cauchy distribution due to the mutual dependence among variables. This implies that the multivariate Cauchy random variable enhances the search capability especially for a large number of correlated variables, and, as a result, is more appropriate for optimization schemes characterized by interdependence among variables. In this sense, the proposed mutation possesses the advantage of both the univariate Cauchy and Gaussian mutations. The proposed mutation is tested against various types of real-valued function optimizations. We empirically find that the proposed mutation outperformed the conventional Cauchy and Gaussian mutations in the optimization of functions having correlations among variables, whereas the conventional mutations showed better performance in functions of uncorrelated variables.

Many kinds of data can be represented as a network or graph. It is crucial to infer the latent structure underlying such a network and to predict unobserved links in the network. Mixed Membership Stochastic Blockmodel (MMSB) is a promising model for network data. Latent variables and unknown parameters in MMSB have been estimated through Bayesian inference with the entire network; however, it is important to estimate them online for evolving networks. In this paper, we first develop online inference methods for MMSB through sequential Monte Carlo methods, also known as particle filters. We then extend them for time-evolving networks, taking into account the temporal dependency of the network structure. We demonstrate through experiments that the time-dependent particle filter outperformed several baselines in terms of prediction performance in an online condition.

Filter caches have been studied as an energy efficient solution. They achieve energy savings via selected access to L1 cache, but severely decrease system performance. Therefore, a filter cache system should adopt components that balance execution delay against energy savings. In this letter, we analyze the legacy filter cache system and propose Data Filter Cache with Partial Tag Cache (DFPC) as a new solution. The proposed DFPC scheme reduces energy consumption of L1 data cache and does not impair system performance at all. Simulation results show that DFPC provides the 46.36% energy savings without any performance loss.

In this paper, we propose a message passing decoding algorithm which lowers decoding error rates in the error floor regions for non-binary low-density parity-check (LDPC) codes transmitted over the binary erasure channel (BEC) and the memoryless binary-input output-symmetric (MBIOS) channels. In the case for the BEC, this decoding algorithm is a combination with belief propagation (BP) decoding and maximum a posteriori (MAP) decoding on zigzag cycles, which cause decoding errors in the error floor region. We show that MAP decoding on the zigzag cycles is realized by means of a message passing algorithm. Moreover, we extend this decoding algorithm to the MBIOS channels. Simulation results demonstrate that the decoding error rates in the error floor regions by the proposed decoding algorithm are lower than those by the BP decoder.

The number of networked devices of sensors and actuators continues to increase. We are developing a data-sharing mechanism called uTupleSpace as middleware for storing and retrieving ubiquitous data that are input or output by such devices. uTupleSpace enables flexible retrieval of sensor data and flexible control of actuator devices, and it simplifies the development of various applications. Though uTupleSpace requires scalability against increasing amounts of ubiquitous data, traditional load-distribution methods using a distributed hash table (DHT) are unsuitable for our case because of the ununiformity of the data. Data are nonuniformly generated at some particular times, in some particular positions, and by some particular devices, and their hash values focus on some particular values. This feature makes it difficult for the traditional methods to sufficiently distribute the load by using the hash values. Therefore, we propose a new load-distribution method using a DHT called the dynamic-help method. The proposed method enables one or more peers to handle loads related to the same hash value redundantly. This makes it possible to handle the large load related to one hash value by distributing the load among peers. Moreover, the proposed method reduces the load caused by dynamic load-redistribution. Evaluation experiments showed that the proposed method achieved sufficient load-distribution even when the load was concentrated on one hash value with low overhead. We also confirmed that the proposed method enabled uTupleSpace to accommodate the increasing load with simple operational rules stably and with economic efficiency.

The hierarchical Dirichlet process (HDP) can provide a nonparametric prior for a mixture model with grouped data, where mixture components are shared across groups. However, the computational cost is generally very high in terms of both time and space complexity. Therefore, developing a method for fast inference of HDP remains a challenge. In this paper, we assume a symmetric multiprocessing (SMP) cluster, which has been widely used in recent years. To speed up the inference on an SMP cluster, we explore hybrid two-level parallelization of the Chinese restaurant franchise sampling scheme for HDP, especially focusing on the application to topic modeling. The methods we developed, Hybrid-AD-HDP and Hybrid-Diff-AD-HDP, make better use of SMP clusters, resulting in faster HDP inference. While the conventional parallel algorithms with a full message-passing interface does not benefit from using SMP clusters due to higher communication costs, the proposed hybrid parallel algorithms have lower communication costs and make better use of the computational resources.

In standard fighting videogames, users usually prefer playing against other users rather than against machines because opponents controlled by machines are in a rut and users can memorize their behaviors after repetitive plays. On the other hand, human players adapt to each other's behaviors, which makes fighting videogames interesting. Thus, in this paper, we propose an artificial agent for a fighting videogame that can adapt to its users, allowing users to enjoy the game even when playing alone. In particular, this work focuses on combination attacks, or combos, that give great damage to the opponent. The agent treats combos independently, i.e., it is composed of a subagent for predicting combos the user executes, that for choosing combos the agent executes, and that for controlling the whole agent. Human users evaluated the agent compared to static opponents, and the agent received minimal negative ratings.

This paper presents a facial expression recognition algorithm based on segmentation of a face image into four facial regions (eyes-eyebrows, forehead, mouth and nose). In order to unify the different results obtained from facial region combinations, a modal value approach that employs the most frequent decision of the classifiers is proposed. The robustness of the algorithm is also evaluated under partial occlusion, using four different types of occlusion (half left/right, eyes and mouth occlusion). The proposed method employs sub-block eigenphases algorithm that uses the phase spectrum and principal component analysis (PCA) for feature vector estimation which is fed to a support vector machine (SVM) for classification. Experimental results show that using modal value approach improves the average recognition rate achieving more than 90% and the performance can be kept high even in the case of partial occlusion by excluding occluded parts in the feature extraction process.

We propose a parameterized multisurface fitting method for multi-frame super-resolution (SR) processing. A parameter assumed for the unknown high-resolution (HR) pixel is used for multisurface fitting. Each surface fitted at each low-resolution (LR) pixel is an expression of the parameter. Final SR result is obtained by fusing the sampling values from these surfaces in the maximum a posteriori fashion. Experimental results demonstrate the superiority of the proposed method.

This paper investigates interference coordination for 3-dimension (3D) antenna array systems in multicell multiple-input multiple-output (MIMO) and orthogonal frequency division multiple-access (OFDMA) wireless networks. Cell-center user and cell-edge user specific downtilts are accordingly partitioned through dynamic vertical beamforming in the 3D MIMO-OFDM communication systems. Taking these user specific downtilts into consideration, the objective of our proposed interference coordination scheme is to maximize both the cell-edge users' and cell-center users' throughput, subject to per base-station (BS) power, cell-center user and cell-edge user specific downtilt constraints. Here, two coordination techniques, consisting of the fractional frequency reuse (FFR) scheme and partial joint process (JP) coordinated multiple point (COMP) transmission mode, are introduced in this paper. To solve the interference coordination problem, two resource block (RB) partitioning schemes are proposed for the above-mentioned coordination techniques accordingly. Based on such RB partitioning, JP CoMP-based dual decomposition method (JC-DDM) and FFR-based dual decomposition method (FDDM) are proposed, where RB assignment, power allocation (RAPA) and downtilts adjustment are jointly optimized. To simplify the computation complexity, a suboptimal algorithm (SOA) is presented to decouple the optimization problem into three subproblems by using FFR scheme. Simulation results show that all of our proposed algorithms outperform the interference coordination scheme with fixed downtilts. JC-DDM and FDDM find the local optimal throughput with different transmission techniques, while SOA iteratively optimize the downtilts and RAPA which shows close-to-optimal performance with much lower computation complexity.

This letter deals with angular position classification using the synthesized active sonar returns from targets. For the synthesis of active sonar returns, we synthesized active sonar returns based on ray tracing algorithm for 3D highlight models. Then, a fractional Fourier transform (FrFT) was applied to the sonar returns to extract the angular position information depending on the target aspect by utilizing separation capability of the time-delayed combination of linear frequency modulated (LFM) signals in the FrFT domain. With the FrFT-based features, three different target angular positions were classified using neural networks.

Compressive sensing is a promising technique in data acquisition field. A central problem in compressive sensing is that for a given sparse signal, we wish to recover it accurately, efficiently and stably from very few measurements. Inspired by mathematical analysis, we introduce a combining scheme between stability and robustness in reconstruction problems using compressive sensing. By choosing appropriate parameters, we are able to construct a condition for reconstruction map to perform properly.

This paper is concerned with a method to reduce the computation time of the Discrete Ray Tracing Method (DRTM) which was proposed to numerically analyze electromagnetic fields above Random Rough Surfaces (RRSs). The essence of DRTM is firstly to search rays between source and receiver and secondly to compute electric fields based on the traced rays. In the DRTM, the method discretizes not only RRSs but also ray tracing procedure. In order to reduce computation time for ray searching, the authors propose to modify the conventional algorithm discretizing RRSs with equal intervals to a new one which discretizes them with unequal intervals according to their profiles. The authors also use an approximation of Fresnel function which enables us to reduce field computation time. The authors discuss the reduction rate for computation time of the DRTM from the numerical view points of ray searching and field computation. Finally, this paper shows how much computation time is reduced by the new method.

Biometric identification has recently attracted attention because of its convenience: it does not require a user ID nor a smart card. However, both the identification error rate and response time increase as the number of enrollees increases. In this paper, we combine a score level fusion scheme and a metric space indexing scheme to improve the accuracy and response time in biometric identification, using only scores as information sources. We firstly propose a score level indexing and fusion framework which can be constructed from the following three schemes: (I) a pseudo-score based indexing scheme, (II) a multi-biometric search scheme, and (III) a score level fusion scheme which handles missing scores. A multi-biometric search scheme can be newly obtained by applying a pseudo-score based indexing scheme to multi-biometric identification. We secondly propose the NBS (Naive Bayes search) scheme as a multi-biometric search scheme and discuss its optimality with respect to the retrieval error rate. We evaluated our proposal using the datasets of multiple fingerprints and face scores from multiple matchers. The results showed that our proposal significantly improved the accuracy of the unimodal biometrics while reducing the average number of score computations in both the datasets.

The single flux quantum (SFQ) is expected to be a next-generation high-speed and low-power technology in the field of logic circuits. CMOS as the dominant technology for conventional processors cannot be replaced with SFQ technology due to the difficulty of implementing feedback loops and conditional branches using SFQ circuits. This paper investigates the applicability of a reconfigurable data-path (RDP) accelerator based on SFQ circuits. The authors introduce detailed specifications of the SFQ-RDP architecture and compare its performance and power/performance ratio with those of a graphics-processing unit (GPU). The results show at most 1600 times higher efficiency in terms of Flops/W (floating-point operations per second/Watt) for some high-performance computing application programs.

A novel analog decoding method using only 90-degree phase shifters is proposed to simplify the decoding method for short-range multiple-input multiple-output (MIMO) transmission. In a short-range MIMO transmission, an optimal element spacing that maximizes the channel capacity exists for a given transmit distance between the transmitter and receiver. We focus on the fact that the weight matrix by zero forcing (ZF) at the optimal element spacing can be obtained by using dividers and 90-degree phase shifters because it can be expressed by a unitary matrix. The channel capacity by the proposed method is next derived for the evaluation of the exact limitation of the channel capacity. Moreover, it is shown that an optimal weight when using directional antennas can be expressed by using only dividers, 90-degree phase shifters, and attenuators, regardless of the beam width of the directional antenna. Finally, bit error rate and channel capacity evaluations by both simulation and measurement confirm the effectiveness of the proposed method.

Let G be a graph with a single source w, assigned a positive integer called the supply. Every vertex other than w is a sink, assigned a nonnegative integer called the demand. Every edge is assigned a positive integer called the capacity. Then a spanning tree T of G is called a spanning distribution tree if the capacity constraint holds when, for every sink v, an amount of flow, equal to the demand of v, is sent from w to v along the path in T between them. The spanning distribution tree problem asks whether a given graph has a spanning distribution tree or not. In the paper, we first observe that the problem is NP-complete even for series-parallel graphs, and then give a pseudo-polynomial time algorithm to solve the problem for a given series-parallel graph G. The computation time is bounded by a polynomial in n and D, where n is the number of vertices in G and D is the sum of all demands in G.

The Discrete Memory Machine (DMM) and the Unified Memory Machine (UMM) are theoretical parallel computing models that capture the essence of the shared memory and the global memory of GPUs. It is assumed that warps (or groups of threads) on the DMM and the UMM work synchronously in a round-robin manner. However, warps work asynchronously in real GPUs, in the sense that they are randomly (or arbitrarily) dispatched for execution. The first contribution of this paper is to introduce asynchronous versions of these models in which warps are arbitrarily dispatched. In addition, we assume that threads can execute the “syncthreads” instruction for barrier synchronization. Since the barrier synchronization operation may be costly, we should evaluate and minimize the number of barrier synchronization operations executed by parallel algorithms. The second contribution of this paper is to show a parallel algorithm to the sum of n numbers in optimal computing time and few barrier synchronization steps. Our parallel algorithm computes the sum of n numbers in O(n/w+llog n) time units and O(log l/log w+log log w) barrier synchronization steps using wl threads on the asynchronous UMM with width w and latency l. Since the computation of the sum takes at least Ω(n/w+llog n) time units, this algorithm is time optimal. Finally, we show that the prefix-sums of n numbers can also be computed in O(n/w+llog n) time units and O(log l/log w+log log w) barrier synchronization steps using wl threads.

Hardware Trojan (HT) has emerged as an impending security threat to hardware systems. However, conventional functional tests fail to detect HT since Trojans are triggered by rare events. Most of the existing side-channel based HT detection techniques just simply compare and analyze circuit's parameters and offer no signal calibration or error correction properties, so they suffer from the challenge and interference of large process variations (PV) and noises in modern nanotechnology which can completely mask Trojan's contribution to the circuit. This paper presents a novel HT detection method based on subspace technique which can detect tiny HT characteristics under large PV and noises. First, we formulate the HT detection problem as a weak signal detection problem, and then we model it as a feature extraction model. After that, we propose a novel subspace HT detection technique based on time domain constrained estimator. It is proved that we can distinguish the weak HT from variations and noises through particular subspace projections and reconstructed clean signal analysis. The reconstructed clean signal of the proposed algorithm can also be used for accurate parameter estimation of circuits, e.g. power estimation. The proposed technique is a general method for related HT detection schemes to eliminate noises and PV. Both simulations on benchmarks and hardware implementation validations on FPGA boards show the effectiveness and high sensitivity of the new HT detection technique.

Methods for in-network joins of sensing data with tuples, in partitioned condition tables stored in sensor nodes, have been studied for efficient event detection. A recently proposed method performs the join operation after distributing the tuples of a condition table evenly among homogeneous sensor nodes with the same storage capacity. In the method, the condition table is horizontally partitioned, and each partition is allocated to the corresponding node, along the path from the highest level to the leaf level. If the path length is larger than the number of partitions, the second round distribution of the partitions resumes from the node at the next level, and so on. Thus, the last node at each round can be assigned the partition that is smaller than the others, which would otherwise cause wasted internal fragmentation. Further, little research has been conducted on methods for the cases of heterogeneous sensor nodes with different available spaces, as well as the vertical partitioning of condition table. In this study, we propose a method of partitioning a condition table that utilizes the internal fragmentation, by treating the tuples of a condition table as a circular list. The proposed method is applicable to the case in which nodes have different available spaces. Furthermore, a new method for vertically partitioning a condition table is suggested. Experiments verify the reduction in the data transmission amount offered by the proposed methods, as compared to existing methods.