Small Memory Capacity Research Articles

Virtual memory partitioning for enhancing application performance in mobile platforms

Recently, the amount of running software on smart mobile devices is gradually increasing due to the introduction of application stores. The application store is a type of digital distribution platform for application software, which is provided as a component of an operating system on a smartphone or tablet. Mobile devices have limited memory capacity and, unlike server and desktop systems, due to their mobility they do not have a memory slot that can expand the memory capacity. Low memory killer (LMK) and out-of-memory killer (OOMK) are widely used memory management solutions in mobile systems. They forcibly terminate applications when the available physical memory becomes insufficient. In addition, before the forced termination, the memory shortage incurs thrashing and fragmentation, thus slowing down application performance. Although the existing page reclamation mechanism is designed to secure available memory, it could seriously degrade user responsiveness due to the thrashing. Memory management is therefore still important especially in mobile devices with small memory capacity. This paper presents a new memory partitioning technique that resolves the deterioration of the existing application life cycle induced by LMK and OOMK. It provides a completely isolated virtual memory node at the operating system level. Evaluation results demonstrate that the proposed method improves application execution time under memory shortage, compared with methods in previous studies.

Computing an upper bound of modularity

Modularity proposed by Newman and Girvan is a quality function for community detection. Numerous heuristics for modularity maximization have been proposed because the problem is NP-hard. However, the accuracy of these heuristics has yet to be properly evaluated because computational experiments typically use large networks whose optimal modularity is unknown. In this study, we propose two powerful methods for computing a nontrivial upper bound of modularity. More precisely, our methods can obtain the optimal value of a linear programming relaxation of the standard integer linear programming for modularity maximization. The first method modifies the traditional row generation approach proposed by Grotschel and Wakabayashi to shorten the computation time. The second method is based on a row and column generation. In this method, we first solve a significantly small subproblem of the linear programming and iteratively add rows and columns. Owing to the speed and memory efficiency of these proposed methods, they are suitable for large networks. In particular, the second method consumes exceedingly small memory capacity, enabling us to compute the optimal value of the linear programming for the Power Grid network (consisting of 4941 vertices and 6594 edges) on a standard desktop computer.

The merit of a small memory capacity: The effect of memory capacity on correlation detection and utilization

本研究では相関の検出と利用における小さな記憶容量のメリットについて，ダミー変数を含む多変数環境を用いて検討した．参加者は各試行において，四つの属性(色，形，大きさ，数)を持つ図形に隠された数字 (1もしくは2) を予測することを求められた．すべての属性は二値変数であり，数字と相関があるのは色のみであった．実験の結果，課題全体を通じてワーキングメモリ容量小群のほうがワーキングメモリ容量大群より課題成績が良かった．したがって，相関の検出や利用において小さなワーキングメモリ容量が有利に働くことが示された．これは小さな記憶容量には適応的な利点があることを示唆する．

MDSA: Modified Distributed Storage Algorithm for Wireless Sensor Networks

In this paper, we propose a Modified distributed storage algorithm for wireless sensor networks (MDSA). Wireless Sensor Networks, as it is well known, suffer of power limitation, small memory capacity,and limited processing capabilities. Therefore, every node may disappear temporarily or permanently from the network due to many different reasons such as battery failure or physical damage. Since every node collects significant data about its region, it is important to find a methodology to recover these data in case of failure of the source node. Distributed storage algorithms provide reliable access to data through the redundancy spread over individual unreliable nodes. The proposed algorithm uses flooding to spread data over the network and unicasting to provide controlled data redundancy through the network. We evaluate the performance of the proposed algorithm through implementation and simulation. We show the results and the performance evaluation of the proposed algorithm.

Discriminant Phase Component for Face Recognition

Numerous face recognition techniques have been developed owing to the growing number of real‐world applications. Most of current algorithms for face recognition involve considerable amount of computations and hence they cannot be used on devices constrained with limited speed and memory. In this paper, we propose a novel solution for efficient face recognition problem for systems that utilize small memory capacities and demand fast performance. The new technique divides the face images into components and finds the discriminant phases of the Fourier transform of these components automatically using the sequential floating forward search method. A thorough study and comprehensive experiments relating time consumption versus system performance are applied to benchmark face image databases. Finally, the proposed technique is compared with other known methods and evaluated through the recognition rate and the computational time, where we achieve a recognition rate of 98.5% with computational time of 6.4 minutes for a database consisting of 2360 images.

Recalibration of Auditory Space following Milliseconds of Cross-Modal Discrepancy

Basic features of objects and events in the environment such as timing and spatial location are encoded by multiple sensory modalities. This redundancy in sensory coding allows recalibration of one sense by other senses if there is a conflict between the sensory maps (Radeau and Bertelson, 1974; Zwiers et al., 2003; Navarra et al., 2009). In contrast to motor or sensorimotor adaptation, which can be relatively rapid, cross-sensory recalibration (the change in an isolated sensory representation after exposure to conflicting cross-modal information) has been reported only as a result of an extensive amount of exposure to sensory discrepancy (e.g., hundreds or thousands of trials, or prolonged durations). Therefore, sensory recalibration has traditionally been associated with compensation for permanent changes that would occur during development or after traumatic injuries or stroke. Nonetheless, the dynamics of sensory recalibration is unknown, and it is unclear whether prolonged inconsistency is required to trigger recalibration or whether such mechanisms are continuously engaged in self-maintenance. We show that in humans recalibration of perceived auditory space by vision can occur after a single exposure to discrepant auditory-visual stimuli lasting only a few milliseconds. These findings suggest an impressive degree of plasticity in a basic perceptual map induced by a cross-modal error signal. Therefore, it appears that modification of sensory maps does not necessarily require accumulation of a substantial amount of evidence of error to be triggered, and is continuously operational. This scheme of sensory recalibration has many advantages. It only requires a small working memory capacity, and allows rapid adaptation to transient changes in the environment as well as the body.

Data reduction method of sine look-up tables in microprocessor's memory storage

A data reduction method of sine look-up tables for low-end microprocessors which have small memory capacity and low process speed is proposed. The proposed method is composed of an encoded sine look-up table (LUT), a linear interpolation technique, and an effective decoding algorithm. The data reduction method minimises the usage of the memory capacity for the sine LUT. It can support low-end microprocessor applications such as single-phase inverter, UPS, direct digital frequency synthesis, etc.

Improved Topic Identification for Similar Document Search on Mobile Devices

This paper presents a novel, two level classifier ensemble designed to support document topic identification in mobile device environments. The proposed system aims at supporting mobile device users who search for documents located in other mobile devices which have similar topic to the documents on the users own device. Conforming to the environment of mobile devices, the algorithms are designed for slower processor, smaller memory capacity and they maintain small data traffic between the devices in order to keep low the cost of communication. We propose a keyword list based topic comparison, enhanced with a two level classifier ensemble to accelerate the topic identification process. The new technique enables document topic comparison using few communication traffic and it requires few calculations.

Learn more by training less: systematicity in sentence processing by recurrent networks

Connectionist models of sentence processing must learn to behave systematically by generalizing from a small training set. To what extent recurrent neural networks manage this generalization task is investigated. In contrast to Van der Velde et al. (Connection Sci., 16, pp. 21–46, 2004), it is found that simple recurrent networks do show so-called weak combinatorial systematicity, although their performance remains limited. It is argued that these limitations arise from overfitting in large networks. Generalization can be improved by increasing the size of the recurrent layer without training its connections, thereby combining a large short-term memory with a small long-term memory capacity. Performance can be improved further by increasing the number of word types in the training set.

광디스크 드라이브용 3축 초박형 액추에이터의 동특성 분석

A note-book PC has become thinner in recent years, which requires the optical disc drives with small height and high memory capacity. Therefore the actuator of optical disc drives must be thinner and have disc tilt compensation function for high density memory. In this paper, the actuator with hybrid type is investigated for 3-axis ultra-slim actuator. A 3-axis ultra-slim actuator is designed by using the modal analysis of the actuator and the electromagnetic analysis of magnetic circuit to achieve dynamic characteristics. Also, magnetic force between tilt magnet and tilt yoke is investigated to and the influence on the DC sensitivity in the focus and track directions.

The neural basis of executive function in working memory: an fMRI study based on individual differences

Using fMRI, neural substrates of the executive system were investigated with respect to differences in working memory capacity. To explore the executive control processes, reading span test (RST) and read conditions were performed. Two subject groups were selected: those with large working memory capacities, labeled high-span subjects (HSS) according to the reading span test, and those with small working memory capacities, labeled low-span subjects (LSS). Significant activation was found mainly in three regions in comparison with the control: anterior cingulate cortex (ACC), left inferior frontal gyrus (IFG), visual association cortex (VAC) and superior parietal lobule (SPL). For both groups, the fMRI signal intensity increased in ACC and IFG during the RST condition compared to that under the read condition. A group difference was also found in the ACC and IFG region, specifically a significant increase in signal intensity was observed only for the HSS group but not for the LSS group. Behavioral data also showed that the performance was better in HSS than in LSS. Moreover, the cross correlation of signal change between ACC and IFG was higher in HSS than in LSS, indicating that the network system between ACC and IFG was more activated in HSS compared to that of LSS. These results suggest that executive function, that is, working attention controlling system is more active in HSS than in LSS. Moreover, the results confirmed our hypothesis that there is a general neural basis for the central executive function in both RST and previous LST (listening span test) tasks despite differences in modality-specific buffers.

The neural basis of individual differences in working memory capacity: an fMRI study

Using fMRI, neural substrates of verbal working memory were investigated with respect to differences in working memory capacity. Listening-span test (LST), Listen, and Remember conditions were performed. Two subjects groups were selected: those who had large working memory capacities, labeled high-span subjects (HSS) according to the working memory span test, and those who had small working memory capacities, labeled low-span subjects (LSS). Significant activation was found mainly in three regions in comparison with resting control: left prefrontal cortex (PFC), anterior cingulate cortex (ACC) and temporal language area. For both groups, fMRI signal intensity increased in PFC during the LST condition compared to the Listen condition. A group difference was found in the ACC region; specifically, a significant increase in signal intensity was observed in ACC only for the HSS group and not for the LSS group. Behavioral data also showed that the performance was better in HSS than in LSS. These results indicate that the attention controlling system, supported by ACC, is more effective in HSS compared to that of LSS.

A Simple Image Coding by Projection of Principal Component in Segmented Color Areas

This paper proposes a simple color image coding method using Principal Component Analysis (PCA) in the segmented color areas. A color image is segmented in the CIELAB color space based on the human perception. In the clustered color area, the chroma a* and b* look to be strongly correlated with lightness L*, and this tendency becomes remarkable by the segmentation. After the segmentation, each cluster is characterized by PCA. The segmented image is labeled with the class number. Since the class number is denoted by an integer value with narrow range but represents the segment color, it is compressed by the conventional loss-less coding. The coded class number is transmitted with the first Principal Component (PC) parameters. Because one set of PC parameters (eigen vector and centroid vector) is transmitted corresponding to each class, it takes small memory capacity. The chroma values of each pixel are approximately reproduced by the projection of first PC onto a*-b* plane along the first PC axis defined by eigen vectors. The first PC value on eigen vector axis mainly carries the lightness information with high resolution. It must be transmitted every pixel and compressed by the conventional Wavelet coding. After restoring the lightness L* from first PC, the full color image is reproduced by combining the estimated chroma components (a*, b*). This paper discusses the coding efficiency and the color reproduction error in relation to the segmented class number.

Optimal allocation of electronic content

The delivery of large files to individual users, such as video on demand or application programs to the envisioned network computers is expected by many to be one of the main tasks of broadband communication networks. This requires high bandwidth capacity as well as fast and dense storage servers. This motivates multimedia service providers to optimize the delivery network, as well as the electronic content allocation. A hierarchical architecture for the distribution of multimedia content was introduced by Nussbaumer, Patel, Schaffa, and Sterbenz (INFOCOM 94). They addressed the trade-off between bandwidth and storage requirements that results from the placement of the content servers in the hierarchy tree. They presented a centralized algorithm to compute the best level of the hierarchy for the server location to minimize the combined cost of communication and storage. In this work, we solve a general case where servers can be placed at different levels of the hierarchy. We develop a distributed optimal location algorithm that requires small nodal memory capacity and computational power. Previous results for related problems for caching system design are of higher complexity. Previous results for related classic operations research problems are limited to centralized algorithms, based on linear programming, that are not easy to convert into distributed algorithms. Instead, to obtain our results, we observed that the use of dynamic programming naturally lends itself to a distributed implementation. For the specific problem at hand, we also managed to find a natural function (a generalization of the problem) that simplifies the combination operation used in the design of a dynamic program.

Analysis of Mixed Convection inside an Automotive Headlamp. 2nd Report. Prediction of the Temperature inside the Parts by FSI-DM Method.

In the former study, in order to predict the temperature in an automotive headlamp without coupling fluid and structure mechanics, SHT (Surface Heat transfer) method has been developed and the temperature estimation accuracy using SHT has been investigated for practical use. Consequently, the temperature results by CFD were within ±10 degrees as compared with the experimental results. The reason of the error was considered with since neither the radiation in the high temperature area nor the three-dimensional movement inside plastic parts were taken into consideration. In the present study, in order to take these movements into consideration, a newly-developed method called FSI-DM (Fluid Solid Interface using Discontinuous Mesh) was developed. This FSI-DM is the method of connecting fluid and solid parts using a discontinuous mesh, and the density of the mesh inside the solid and the fluid can be changeable. Furthermore, the coarse mesh on the surface of the solid enables us to calculate the radiation with a smaller memory capacity.

Facing the challenge of wireless security

Increasingly, companies and individuals are using wireless technology for important communications they want to keep private, such as mobile e-commerce transactions, email, and corporate data transmissions. At the same time, as wireless platforms mature, grow in popularity, and store valuable information, hackers are stepping up their attacks on these new targets. This is a particular problem because wireless devices, including smart cellular phones and personal digital assistants (PDAs) with Internet access, were not originally designed with security as a top priority. Now, however, wireless security is becoming an important area of product research and development. As in the wired world, wireless security boils down to protecting information and preventing unauthorized system access. However, it is challenging to implement security in small-footprint devices with low processing power and small memory capacities and that use unreliable, low bandwidth wireless networks. Vendors and others have developed several security approaches for the various wireless technologies, although each of these early efforts has some shortcomings. Security researchers are thus busy developing new technologies and fixing holes in existing ones. The paper discusses wireless security technology and its shortcomings.

Fault tolerance of reconfigurable bi-directional double-loop LANs

Double-loop is a very popular structure in loop network topology. A reconfigurable bi-directional double-loop structure is recently developed. It has a new structure which is different from any existing double-loop structure and no previous methods can be used to evaluate its fault tolerance which is demanded for all fresh ideas. This paper provides an easy approach to offset this deficiency. A network is segmented first based on its block connection and then analyzed in a recursive way. The calculation using this method requires linear time and very small memory capacity. This method can be used in analyzing the fault tolerance of other topologically segmental loop networks.

Rigid-plastic finite-element method for analysis of three-dimensional rolling that requires small memory capacity

At present, caliber rolling is analyzed using the three-dimensional rigid-plastic finite-element method. In this paper, a new method of analysis which requires less memory capacity than the conventional method is proposed. In the conventional method, the shape of the whole region is updated after the functional of the region is minimized, whereas in the proposed method, the shape of part of the region (large region) is updated after the functional of the large region is minimized. The load, torque, material shape, hydrostatic stress and equivalent strain rate obtained using the proposed method agree with those obtained using the conventional method within the allowable range of convergent error. Also, although the computation time for the proposed method is rather long compared with that for the conventional method, the memory capacity required for the proposed method is small compared with that for the conventional method. Finally, analysis of various types of caliber rolling in which the number of finite elements in the large region coincides with that in the whole region in analysis using the conventional method is carried out and reasonable stress and strain distributions are obtained.

Method and apparatus for speech processing

An apparatus and method for processing vocal information includes an extractor for extracting a plurality of spectrum information from parameters for vocal information, a vector quantizer for vector-quantizing the extracted spectrum information and for producing a plurality of parameter patterns therefrom, a memory for storing the plurality of parameter patterns so obtained, and a memory for storing positional information indicating the positions at which the plurality of parameter patterns are stored and for storing code information specifying parameter patterns and corresponding to the positional information. The parameter patterns and code information can be used to synthesize speech. Because a small number of parameter patterns are used, only a small memory capacity is needed and efficient processing of vocal information can be performed.

A personal Computer Aided Inverse Finite Element Analysis. An Efficient Identification Method of Unknown Boundary Condition by Penalty Function Method.

This paper presents an efficient inverse analysis technique for the identification of an unknown boundary condition in a finite element method (FEM) on a small memory capacity machine such as a personal computer. The author has previously presented the FEM of inverse analysis introducing an inverse compliance matrix that uses an influence function into the stiffness matrix for the unknown boundary. Since the stiffness matrix of the FEM becomes singular, following two calculation processes are adopted: (1) a calculation process which obtains an approximate solution by decreasing the effect of the inverse compliance in the stiffness, and (2) an iterative calculation process computing a correct solution from the approximation using (1). First, a penalty function method is used to decrease the rank of a singular stiffness matrix in (1). Then, the number for computing LDU decomposed matrix becomes twice in comparison with fourth times in the previous paper. Next, an efficient calculation method is presented using an attack technique based on both accelerative moving lower and upper bounds in (2). Moreover, a zooming identification method is developed, owing to an improvement of node displacements on and near the edges. Therefore, this technique makes it possible to establish an efficient inverse analysis for the identification of the unknown boundary. Finally, the validity of the inverse analysis technique is confirmed from the numerical solution of a three-dimensional inverse finite element analysis model, as an example of an elastic structure with reinforced members under tension, carried out on a small memory capacity machine.