Kernel methods such as the support vector machine are one of the most successful algorithms in modern machine learning. Their advantage is that linear algorithms are extended to non-linear scenarios in a straightforward way by the use of the kernel trick. However, naive use of kernel methods is computationally expensive since the computational complexity typically scales cubically with respect to the number of training samples. In this article, we review recent advances in the kernel methods, with emphasis on scalability for massive problems.

View full abstract

Particle Swarm Optimization (PSO) is a search method which utilizes a set of agents that move through the search space to find the global minimum of an objective function. The trajectory of each particle is determined by a simple rule incorporating the current particle velocity and exploration histories of the particle and its neighbors. Since its introduction by Kennedy and Eberhart in 1995, PSO has attracted many researchers due to its search efficiency even for a high dimensional objective function with multiple local optima. The dynamics of PSO search has been investigated and numerous variants for improvements have been proposed. This paper reviews the progress of PSO research so far, and the recent achievements for application to large-scale optimization problems.

View full abstract

The spike timings of neurons are irregular and are considered to be a one-dimensional point process. The Bayesian approach is generally used to estimate the time-dependent firing rate function from sequences of spike timings. It can also be used to estimate the firing rate from only a single sequence of spikes. However, the rate function has too many degrees of freedom in general, so approximation techniques are often used to carry out the Bayesian estimation. We applied the transfer matrix method, which efficiently computes the exact marginal distribution, to the estimation of the firing rate and developed an algorithm that enables the exact results to be obtained for the Bayesian framework. Using this estimation method, we investigated how the mismatch of the prior hyperparameter value affects the marginal distribution and the firing rate estimation.

View full abstract

Traditional frequent pattern mining algorithms do not consider different semantic significances (weights) of the items. By considering different weights of the items, weighted frequent pattern (WFP) mining becomes an important research issue in data mining and knowledge discovery area. However, the existing state-of-the-art WFP mining algorithms consider all the data from the very beginning of a database to discover the resultant weighted frequent patterns. Therefore, their approaches may not be suitable for the large-scale data environment such as data streams where the volume of data is huge and unbounded. Moreover, they cannot extract the recent change of knowledge in a data stream adaptively by considering the old information which may not be interesting in the current time period. Another major limitation of the existing algorithms is to scan a database multiple times for finding the resultant weighted frequent patterns. In this paper, we propose a novel large-scale algorithm WFPMDS (Weighted Frequent Pattern Mining over Data Streams) for sliding window-based WFP mining over data streams. By using a single scan of data stream, the WFPMDS algorithm can discover important knowledge from the recent data elements. Extensive performance analyses show that our proposed algorithm is very efficient for sliding window-based WFP mining over data streams.

View full abstract

The D_st index is the most popular measure of a scale of magnetic storms, and it is widely used as a monitor of the conditions of the Earth's magnetosphere. Since the D_st index contains contributions from multiple magnetospheric phenomena, it is important to distinguish each of the contributions in order to obtain meaningful information about the conditions of the magnetosphere. There have been several efforts which modeled temporal evolution of the D_st index empirically, and these empirical models considers some contributions separately. However, they take only short-term varations into accout, and contributions from phenomena which show long-term variations are neglected. In the present study, we have developed a technique for estimating the component of long-term variations of the D_st index using solar wind data and a nonlinear empirical model. The newly-developed technique adopts an algorithm which is similar to the particle filter. This algorithm allows an on-line processing of a long sequence of D_st data, which would enable a real-time estimation of system variables in a nonlinear system model. The estimates of the long-term variations can be used for accurate estimation of other contributions to the D_st index, which would provide credible information about the conditions of the magnetosphere. The framework proposed in the present study could be applied for the purpose of continuous real-time monitoring of the environment of the magnetosphere.

View full abstract

This paper presents an efficient algorithm for large-scale multi-system learning task. The proposed architecture, referred to as the ‘RBFxSOM’, is based on the SOM², that is, a‘SOM of SOMs’. As is the case in the modular network SOM (mnSOM) with multilayer perceptron modules (MLP-mnSOM), the aim of the RBFxSOM is to organize a continuous map of nonlinear functions representing multi-class input-output relations of the given datasets. By adopting the algorithm for the SOM², the RBFxSOM generates a map much faster than the original mnSOM, and without the local minima problem. In addition, the RBFxSOM can be applied to more difficult cases, that were not easily dealt with by the MLP-mnSOM. Thus, the RBFxSOM can deal with cases in which the probability density of the inputs is dependent on the classes. This tends to happen more often as the input dimension increases. The RBFxSOM therefore, overcomes many of the problems inherent in the MLP-mnSOM, and this is crucial for application to large scale tasks. Simulation results with artificial datasets and a meteorological dataset confirm the performance of the RBFxSOM.

View full abstract

With a Gaussian kernel function, we find that the distance between two classes (DBTC) can be used as a class separability criterion in feature space since the between-class separation and the within-class data distribution are taken into account impliedly. To test the validity of DBTC, we develop a method of tuning the kernel parameters in support vector machine (SVM) algorithm by maximizing the DBTC in feature space. Experimental results on the real-world data show that the proposed method consistently outperforms corresponding hyperparameters tuning methods.

View full abstract

We propose a user-defined on-demand matching strategy, called O-matching, in which users can control the order of matching arguments of each operation symbol. In ordinary matching schemes it is not important to set the order of matching, however, in on-demand matching schemes, it is very important since an input term may be changed while doing the on-demand matching process. O-matching is suitable to combine with the E-strategy, which is a user-defined reduction strategy in which users can control the order of reducing arguments. We show a sufficient condition under which the E-strategy with O-matching is correct for head normal forms, that is, any reduced term is a head normal form.

View full abstract

A task that suspends itself to wait for an I/O completion or to wait for an event from another node in distributed environments is called an I/O blocking task. Conventional hard real-time scheduling theories use framework of rate monotonic analysis (RMA) to schedule such I/O blocking tasks. However, most of them are pessimistic. In this paper, we propose effective algorithms that can schedule a task set which has I/O blocking tasks under dynamic priority assignment. We present a new critical instant theorem for the multi-frame task set under dynamic priority assignment. The schedulability is analyzed under the new critical instant theorem. For the schedulability analysis, this paper presents saturation summation which is used to calculate the maximum interference function (MIF). With saturation summation, the schedulability of a task set having I/O blocking tasks can be analyzed more accurately. We propose an algorithm which is called Frame Laxity Monotonic Scheduling (FLMS). A genetic algorithm (GA) is also applied. From our experiments, we can conclude that FLMS can significantly reduce the calculation time, and GA can improve task schedulability ratio more than is possible with FLMS.

View full abstract

A data stream is a series of massive unbounded tuples continuously generated at a rapid rate. Continuous queries for data streams should be processed continuously, so that a strict time constraint is required. In most previous research studies, in order to guarantee this constraint, the evaluation order of join predicates in a continuous query is optimized using a greedy strategy. However, because a greedy strategy traces only the first promising plan, it often finds a suboptimal plan. To reduce the possibility of producing a suboptimal plan, in this paper, we propose an improved scheme, k-Extended Greedy Algorithm (k-EGA), that simultaneously examines a set of promising plans and reoptimize an execution plan adaptively. The number of promising plans is flexibly controlled by a user-defined range variable. The scheme verifies the performance of the current plan periodically. If the plan is no longer efficient, a newly optimized plan is generated. The performance of the proposed scheme is verified through various experiments to identify its various characteristics.

View full abstract

A multiple stream join is one of the most important but high cost operations in ubiquitous streaming services. In this paper, we propose a newly improved and practical algorithm for joining multiple streams called AMJoin, which improves the multiple join performance by guaranteeing the detection of join failures in constant time. To achieve this goal, we first design a new data structure called BiHT (Bit-vector Hash Table) and present the overall behavior of AMJoin in detail. In addition, we show various experimental results and their analyses for clarifying its efficiency and practicability.

View full abstract

The ITU-T J.83 Annex B is a widely adopted standard in North America for digital video and audio transmission over coaxial cable. This paper proposes a new parallel processing architecture of the parity checksum generator and syndrome generator specified in the standard for packet synchronization and error detection. The proposed parallel processing architecture removes the performance bottleneck occurring in the conventional serial processing architecture, leading to significant decrease in processing time for generating a parity checksum in transmitter and a syndrome in receiver. Implementation results show that the proposed parallel processing architecture reduces the processing time by 92% for parity checksum generation and by 81% for syndrome generation over the conventional serial processing architecture.

View full abstract

We present a real-time video-based rendering system using a network camera array. Our system consists of 64 commodity network cameras that are connected to a single PC through a gigabit Ethernet. To render a high-quality novel view, our system estimates a view-dependent per-pixel depth map in real time by using a layered representation. The rendering algorithm is fully implemented on the GPU, which allows our system to efficiently perform capturing and rendering processes as a pipeline by using the CPU and GPU independently. Using QVGA input video resolution, our system renders a free-viewpoint video at up to 30 frames per second, depending on the output video resolution and the number of depth layers. Experimental results show high-quality images synthesized from various scenes.

View full abstract

This paper proposes an approach to image segmentation using Iterated Graph Cuts based on local texture features of wavelet coefficients. Using Haar Wavelet based Multiresolution Analysis, the low-frequency range (smoothed image) is used for the n-link and the high-frequency range (local texture features) is used for the t-link along with the color histogram. The proposed method can segment an object region having not only noisy edges and colors similar to the background, but also heavy texture change. Experimental results illustrate the validity of our method.

View full abstract

This paper presents a grouped scan slice encoding technique using scan slice repetition to simultaneously reduce test data volume and test application time. Using this method, many scan slices that would be incompatible with the conventional selective scan slice method can be encoded as compatible scan slices. Experiments were performed with ISCAS'89 and ITC'99 benchmark circuits, and results show the effectiveness of the proposed method.

View full abstract

We present a numerical analysis of the optimal number of active mobile users for minimizing average data delivery delay in intelligent people-centric urban sensing, in which context-aware mobile devices act as sensor-data carriers and sensor nodes act as data accumulators within CDMA cellular networks. In the analysis, we compute the optimal number of mobile users for different environmental conditions and then investigate the minimum average data delivery delay for this optimal number of mobile users.

View full abstract

This letter proposes a busy-tone based scheme for reliable and efficient broadcasting in mobile ad hoc networks. Control packets such as RTS, CTS and ACK are ignored in the broadcast scheme, and two busy tones are used, one for channel reservation and the other for negative acknowledgement. Unlike traditional schemes for reliable broadcasting, the proposed scheme is highly efficient as it achieves both collision avoidance and fast packet loss recovery. Simulation results are presented which show the effectiveness of the proposed scheme.

View full abstract

Most large-scale distributed file systems decouple a metadata operation from read and write operations for a file. In the distributed file systems, a certain server named a metadata server (MDS) is responsible for maintaining the metadata information of the file systems. In this paper, we propose a new metadata management scheme in order to provide the high metadata throughput and scalability for a cluster of MDSs. First, we derive a new metadata distribution technique. Then, we present a load balancing technique based on the distribution technique. Several experiments show that our scheme outperforms existing metadata management scheme in terms of scalability and load balancing.

View full abstract

In a large-scale sensor network, replicated hostile nodes may be used for harsh inner attacks. To detect replicas, this paper presents a distributed, deterministic, and efficient approach robust to node compromise attacks without incurring significant resource overheads.

View full abstract

In this paper, we propose block matching and learning for color image classification. In our method, training images are partitioned into small blocks. Given a test image, it is also partitioned into small blocks, and mean-blocks corresponding to each test block are calculated with neighbor training blocks. Our method classifies a test image into the class that has the shortest total sum of distances between mean blocks and test ones. We also propose a learning method for reducing memory requirement. Experimental results show that our classification outperforms other classifiers such as support vector machine with bag of keypoints.

View full abstract

A hidden Markov model (HMM)-based parameter estimation scheme is proposed for wideband speech recovery. In each Markov state, the estimation efficiency is improved using a new mapping function derived from the weighted least squares of vector deviations. The experimental results reveal that the performance of the proposed scheme is superior to that combining the HMM and Gaussian mixture model (GMM).

View full abstract

In this letter, an acoustic environment classification algorithm based on the 3GPP2 selectable mode vocoder (SMV) is proposed for context-aware mobile phones. Classification of the acoustic environment is performed based on a Gaussian mixture model (GMM) using coding parameters of the SMV extracted directly from the encoding process of the acoustic input data in the mobile phone. Experimental results show that the proposed environment classification algorithm provides superior performance over a conventional method in various acoustic environments.

View full abstract

Moving object completion is a process of completing moving object's missing information based on local structures. Over the past few years, a number of computable algorithms of video completion have been developed, however most of these algorithms are based on the pixel domain. Little theoretical and computational work in video completion is based on the compressed domain. In this paper, a moving object completion method on the compressed domain is proposed. It is composed of three steps: motion field transferring, thin plate spline interpolation and combination. Missing space-time blocks will be completed by placing new motion vectors on them so that the resulting video sequence will have as much global visual coherence with the video portions outside the hole. The experimental results are presented to demonstrate the efficiency and accuracy of the proposed algorithm.

View full abstract

We propose an optimized scanline filling algorithm for OpenVGtwo-dimensional vector graphics. For each scanline of a path, it adaptively selects a left or right scanning direction that minimizes the number of pixels visited during scanning. According to the experimental results, the proposed algorithm reduces the number of pixels visited by 6 to 37% relative to that with a constant scanning direction for all the scanlines.

View full abstract

A method using genetic algorithms for path generation have been proposed; however, this method is limited to particular applications, and there are limitations on the types of paths that can be represented. This paper therefore proposes a path generation method that is applicable to more general-purpose applications compared to previous methods based on a new design of the genotype used in the genetic algorithm.

View full abstract

The fuzzy controller based on cycle-to-cycle control with output value adjustment factors (OAF) was developed for restoring gait of paralyzed subjects by using functional electrical stimulation (FES). Results of maximum knee flexion and extension controls with neurologically intact subjects suggested that the OAFs would be effective in reaching the target within small number of cycles and in reducing the error after reaching the target. Oscillating responses between cycles were also suppressed. The fuzzy controller was expected to be examined to optimize the OAFs with more subjects including paralyzed patients for clinical application.

View full abstract