Memory Reduction Research Articles

Sparse matrices for transient simulations with computing memory reduction

Modified numerical integration methods and modelling have been analysed for application to simulations of electromagnetic transient phenomena in transmission lines. Both objectives are considered: minimize numerical errors caused by Gibbs’ oscillations and decrease the computational memory necessary for the mentioned simulations. For this, transmission lines have represented by cascades of π circuits. Based on this representation, modifications in numerical integration methods applied to the mentioned simulations can be introduced easily and the comparisons can be carried out quickly. In case of minimization of numerical errors, two approaches have checked. One of these options is related to the way how the numerical integration method is applied. The numerical integration has applied using 2-order matrices. These matrices are related to each π circuit applied to represent transmission lines. The other option is related to the damping resistance application. It has made by introducing damping resistances in the longitudinal structure of the π circuits. Sparse matrices have applied for decreasing the computational memory used during the simulations, because this matrix has a great quantity of null elements when the cascade of π circuits are represented by only one matrix with great order. Comparisons have carried out considering the non-modified numerical methods and the three modified methods suggested in this paper.

Nuclear Fuel Assembly Deformation, Reduced Mechanical Model Dedicated to FSI Simulation

In the pressure vessel of a PWR the core is constituted of a number of fuel assemblies. Under operation, they all suffer from various phenomena, among them, initial trapped efforts, thermal expansion, irradiation growth, creeping, tightening relaxation and rod slipping. Each fuel assembly may develop contacts with its neighbors and strongly interacts with the fluid. It then undergoes vertical as well as lateral fluid forces, which are partly coupled with its instantaneous shape. As a consequence the fuel assembly tends to get distorted during and still after operation. The first concern is economic: at the end of a cycle, when changing the assembly position in the core, the contact interactions with its neighbors due to its distorted shape sometimes delay the handling. The second concern is that the shutdown rods have to slide into the deformed shape of the guide tube and might be slowed due to excessive friction forces. In order to precisely assess these situations it is necessary to simulate numerically a complete core, taking into account all of the previously mentioned phenomena. For that purpose, we propose a light but relevant reduced mechanical model of a fuel assembly, partially based on a POD analysis performed on a detailed fuel assembly model. We then show that a rather simple model, still based on classical finite elements can be used and simplifies the integration of the local nonlinear constitutive equations. We then compare both of the detailed and reduced model on typical situations and show excellent agreement. The CPU time ratio between them, for a full nonlinear simulation of a single fuel assembly under operation, is over 1000, which is competitive against separation of variables based reduced order methods such as POD, PGD and even APHR. So, yet still with a simplified FSI approach, the drastic memory and CPU reductions obtained now enable us to complete a 3D core simulation representing at least 4 years under operation (4 cycles), with as much as 241 individual fuel assemblies, in a reasonable time of a few hours.

Systematic Monitoring of Cognitive Function During Electroconvulsive Therapy: A Retrospective Analysis of Data From a Service Using a Short Cognitive Testing Battery.

International guidelines suggest repeating cognitive testing at intervals throughout a course of electroconvulsive therapy (ECT) to monitor its effects on cognitive function. However, it is apparent that few services do this, and an optimal battery of testing has not yet been established. We aimed to evaluate the utility of such routine cognitive testing in a clinic where patients had been routinely tested at intervals throughout a course of ECT. All patients referred for ECT at a public ECT clinic were offered routine cognitive testing to monitor cognitive function during their course of ECT. Testing was conducted at baseline and after 3, 6, and 9 treatments. Analyses examined whether change in individual measures predicted reduction in autobiographical memory at subsequent measures and whether the results that were given to clinicians informed treatment decisions. Changes in cognitive test results were not associated with clinician decisions to change treatment parameters. Only change in digit span forwards after 3 treatments was associated with later reduction in Colombia University Autobiographical Interview - Short Form (CUAMI-SF) of greater than 25%, with a larger improvement in digit span forwards being associated with greater chance of having a 25% reduction in CUAMI-SF. There was no evidence that the screening undertaken in this clinic had been helpful in determining treatment decisions or that changes in cognitive tests predicted in a reliable way who would later experience changes in autobiographical memory. However, follow-up testing was not completed reliably, and longer-term data regarding autobiographical memory were not collected.

A quantum solution for efficient use of symmetries in the simulation of many-body systems

A many-body Hamiltonian can be block-diagonalized by expressing it in terms of symmetry-adapted basis states. Finding the group orbit representatives of these basis states and their corresponding symmetries is currently a memory/computational bottleneck on classical computers during exact diagonalization. We apply Grover’s search in the form of a minimization procedure to solve this problem. Our quantum solution provides an exponential reduction in memory, and a quadratic speedup in time over classical methods. We discuss explicitly the full circuit implementation of Grover minimization as applied to this problem, finding that the oracle only scales as polylog in the size of the group, which acts as the search space. Further, we design an error mitigation scheme that, with no additional qubits, reduces the impact of bit-flip errors on the computation, with the magnitude of mitigation directly correlated with the error rate, improving the utility of the algorithm in the Noisy Intermediate Scale Quantum era.

Identification of Weather Phenomena Based on Lightweight Convolutional Neural Networks

Weather phenomenon recognition plays an important role in the field of meteorology. Nowadays, weather radars and weathers sensor have been widely used for weather recognition. However, given the high cost in deploying and maintaining the devices, it is difficult to apply them to intensive weather phenomenon recognition. Moreover, advanced machine learning models such as Convolutional Neural Networks (CNNs) have shown a lot of promise in meteorology, but these models also require intensive computation and large memory, which make it difficult to use them in reality. In practice, lightweight models are often used to solve such problems. However, lightweight models often result in significant performance losses. To this end, after taking a deep dive into a large number of lightweight models and summarizing their shortcomings, we propose a novel lightweight CNNs model which is constructed based on new building blocks. The experimental results show that the model proposed in this paper has comparable performance with the mainstream non-lightweight model while also saving 25 times of memory consumption. Such memory reduction is even better than that of existing lightweight models.

Optimization of Software on High Performance Computing Platforms for the LUX-ZEPLIN Dark Matter Experiment

High Energy Physics experiments like the LUX-ZEPLIN dark matter experiment face unique challenges when running their computation on High Performance Computing resources. In this paper, we describe some strategies to optimize memory usage of simulation codes with the help of profiling tools. We employed this approach and achieved memory reduction of 10-30%. While this has been performed in the context of the LZ experiment, it has wider applicability to other HEP experimental codes that face these challenges on modern computer architectures.

Approximation-Aware Design of an Image-Based Control System

Image-based control (IBC) systems are common in many modern applications. In such systems, image-based sensing imposes massive compute workload, making them challenging to implement on embedded platforms. Approximate image processing is a way to handle this challenge. In essence, approximation reduces the workload at the cost of additional sensor noise. In this work, we propose an approximation-aware design approach for optimizing the energy, memory and performance of an IBC system, making it suitable for embedded implementation. First, we perform compute- and data-centric approximations and evaluate its impact on the energy efficiency, memory utilization and closed-loop quality-of-control (QoC) of the IBC system. We observe that the workload reductions due to approximations allow mapping these lighter approximated IBC tasks to embedded platforms with lower power consumption while still ensuring proper system functionality. Therefore, we explore the interplay between approximations and platform mappings to improve the energy-efficiency of IBC systems. Further, an IBC system operates under several environmental scenarios e.g., weather conditions. We evaluate the sensitivity of the IBC system to our approximation-aware design approach when operated under different scenarios and perform a failure probability (FP) analysis using Monte-Carlo simulations to analyze the robustness of the approximate system. Finally, we design an optimal approximation-aware controller that models the approximation error as sensor noise and show QoC improvements. We demonstrate the effectiveness of our approach using a concrete case-study of a lane keeping assist system (LKAS) using a heterogeneous NVIDIA AGX Xavier embedded platform in a hardware-in-the-loop (HiL) framework. We show energy and memory reduction of up to 92% and 88% respectively, for 44% QoC improvements with respect to the accurate implementation. We show that our approximation-aware design approach has an FP (per km) $\leq 9.6\times 10^{-6}$ %.

Guaranteed memory reduction in synthesis of correct-by-design invariance controllers

Formal methods for analysis of dynamical systems through construction of finite symbolic abstractions have attracted significant interest as they allow solving complex control problems in a fully automated fashion. Nevertheless, their practical application is currently limited by the fact that they require enormous memory resources. We present a novel algorithm for solution of invariance problems within abstraction-based framework, which guarantees large storage reduction and fully applies to general non-linear plants. We also show that, in practice, the algorithm is faster compared to other methods.

Reduction in Space for Dynamic Finite Element Analysis of Assemblies of Beam-columns when the Mass is Available in Digitized Format

In 2008, a technique was proposed to reduce run-times in analysis of semi-discretized equation of motion against dynamic excitations available in digitized format. Later, the technique was successfully adapted to reduce numbers of degrees of freedom in finite element analysis of assemblies of beam-columns subjected to static digitized loads. In this paper, attention is paid to dynamic finite element analysis of assemblies of beam-columns. It is shown that, when the mass is available in digitized format, after small modifications in the original technique, the adaptation can simplify the analysis, regardless of the models' sizes, their linearity or non-linearity, and whether the damping is classical or non-classical. The reductions in run-time and in-core memory are considerable, while the changes in accuracy can be negligible.

Early Postnatal Ethanol Exposure in Mice Induces Sex-Dependent Memory Impairment and Reduction of Hippocampal NMDA-R2B Expression in Adulthood

Drinking alcohol during pregnancy is particularly detrimental for the developing brain and may cause a broad spectrum of cognitive and behavioral impairments, collectively known as fetal alcohol spectrum disorder (FASD). While behavioral abnormalities and brain damage have been widely investigated in animal models of FASD, the sex differences in the vulnerability to perinatal ethanol exposure have received less consideration. Here we investigated the long-term behavioral and molecular effects of acute ethanol-binge like exposure during the early postnatal period (equivalent to the third trimester of human pregnancy) in adult male and female mice. CD1 mice received a single ethanol exposure on P7 and were analyzed starting from P60. We found that ethanol-exposed mice showed increased activity in the open field test and in the plus-maze test, regardless of the sex. Interestingly, only ethanol-exposed adult male mice exhibited memory impairment in the water maze and fear-conditioning tests. Remarkably, hippocampal levels of NMDA-R2B were reduced only in ethanol-exposed male, while total BDNF levels were increased in both male and female ethanol-exposed mice.Our data suggest a different susceptibility of early postnatal ethanol exposure in male and female CD1 mice.

Memory for actions in Chinese children with developmental dyslexia.

Around 5-10% of school children in China suffer from reading disabilities. Previous studies have shown that children with developmental dyslexia have deficits in encoding and retrieval information. Moreover, many studies have provided evidence that performing actions can promote the encoding and retrieval of memory compared to verbal tasks. Therefore, we expected that performing actions can improve the memory performance of children with developmental dyslexia. However, memory for actions of children with developmental dyslexia has not been sufficiently explored. The present study used a 2 (group: dyslexic or typical) × 2 (type of encoding: subject-performed task or verbal task) between-subjects design to examine whether subject-performed task could enhance the memory performance of children with developmental dyslexia. The results revealed that performing actions not only improves the memory performance of the typical group but also of the dyslexic group. We suggest that performing actions can improve the level of self-involvement during encoding and enhance item accessibility during retrieval of memory of children with dyslexia, thus compensating for the memory reduction caused by the deficits of rehearsal and retrieval strategies of children with dyslexia.

Fast and Accurate Depth Estimation from Sparse Light Fields.

We present a fast and accurate method for dense depth reconstruction, which is specifically tailored to process sparse, wide-baseline light field data captured with camera arrays. In our method, the source images are over-segmented into non-overlapping compact superpixels. We model superpixel as planar patches in the image space and use them as basic primitives for depth estimation. Such superpixel-based representation yields desired reduction in both memory and computation requirements while preserving image geometry with respect to the object contours. The initial depth maps, obtained by plane-sweeping independently for each view, are jointly refined via iterative belief-propagation-like optimization in superpixel domain. During the optimization, smoothness between the neighboring superpixels and geometric consistency between the views are enforced. To ensure rapid information propagation into textureless and occluded regions, together with the immediate superpixel neighbors, candidates from larger neighborhoods are sampled. Additionally, in order to make full use of the parallel graphics hardware a synchronous message update schedule is employed allowing to process all the superpixels of all the images at once. This way, the distribution of the scene geometry becomes distinctive already after the first iterations, facilitating stability and fast convergence of the refinement procedure. We demonstrate that a few refinement iterations result in globally consistent dense depth maps even in the presence of wide textureless regions and occlusions. The experiments show that while the depth reconstruction takes about a second per full high-definition view, the accuracy of the obtained depth maps is comparable with the state-of-the-art results, which otherwise require much longer processing time.

Когнітивні порушення у осіб з алкогольною енцефалопатією та пароксизмальними станами

article addresses the issues of cognitive impairment among persons with alcohol dependence, aggravated by alcoholic encephalopathy and paroxysmal conditions.The aim of the study was to identify the levels of impaired short-term memory, to identify cognitive impairment of patients with alcohol addiction with alcoholic encephalopathy and paroxysmal conditions.Materials and methods: 132 people from the contingent of alcohol addicts (AА) and from the contingent of healthy and 4 comparison groups have been identified and examined over the two years on the basis of KNP CHOR «Regional Clinical Narcological Hospital No. 3». following techniques were used to assess psychosocial and cognitive impairment: “Jacobson Short-Term Memory Measurement Technique”; The methodology for determining the index of short-term memory proposed by L. S. Muchnik and V. M. Smirnov (1968).Result. According to the results of the researches, the cognitive and mnemonic sphere of persons with AА, AE and PS were expressed in the form of significant reduction of short-term memory and cognitive impairment. presence of significantly deeper and gross degenerative-organic lesions of the central nervous system in chronic alcoholic lesions with the development of alcoholic encephalopathy and paroxysmal states of alcoholic genesis has been proved.Conclusions. Thus, the obtained research results only confirm the data of numerous world sources on the development of cognitive decline in individuals with alcohol dependence

β-Amyloid Peptide: the Cell Compartment Multi-faceted Interaction in Alzheimer's Disease.

Alzheimer's disease (AD) is the most widespread form of dementia, characterized by memory loss and reduction of cognitive functions that strongly interfere with normal daily life. Numerous evidences show that aggregates of the amyloid beta peptide, formed by 39 to 42 amino acid residues (Aβ39-43), from soluble small oligomers to large fibrils are characteristic markers of this pathology. However, AD is a complex disease and its neurodegenerative molecular mechanism is not yet fully understood. Growing evidence suggests a link between Aβ polymorphic nature, oligomers and fibrils, and specific mechanisms of neurodegeneration. The Aβ variable nature and its multiplicity of interactions with different proteins and organelles reflect the complexity of this pathology. In this review, we analyze the effects of the interaction between Aβ peptide and different cellular compartments in relation to the different kinds and sizes of amyloid aggregates. In particular, Aβ interaction with different cell structures such as the plasma membrane, mitochondria, lysosomes, nucleus, and endoplasmic reticulum is discussed. Further, we analyze the Aβ peptide ability to modify the structure and function of the target organelle, inducing alteration of its physiological role thus contributing to the pathological event. Dysfunction of cellular components terminating with the activation of the cellular death mechanism and subsequent neurodegeneration is also taken into consideration.

Parallelized collision detection with applications in virtual bone machining

Background and objectivesVirtual reality surgery simulators have been proved effective for training in several surgical disciplines. Nevertheless, this technology is presently underutilized in orthopaedics, especially for bone machining procedures, due to the limited realism in haptic simulation of bone interactions. Collision detection is an integral part of surgery simulators and its accuracy and computational efficiency play a determinant role on the fidelity of simulations. To address this, the primary objective of this study was to develop a new algorithm that enables faster and more accurate collision detection within 1 ms (required for stable haptic rendering) in order to facilitate the improvement of the realism of virtual bone machining procedures. MethodsThe core of the developed algorithm is constituted by voxmap point shell method according to which tool and osseous tissue geometries were sampled by points and voxels, respectively. The algorithm projects tool sampling points into the voxmap coordinates and compute an intersection condition for each point-voxel pair. This step is massively parallelized using Graphical Processing Units and it is further accelerated by an early culling of the unnecessary threads as instructed by the rapid estimation of the possible intersection volume. A contiguous array was used for implicit definition of voxmap in order to guarantee a fast access to voxels and thereby enable efficient material removal. A sparse representation of tool points was employed for efficient memory reductions. The effectiveness of the algorithm was evaluated at various bone sampling resolutions and was compared with prior relevant implementations. ResultsThe results obtained with an average hardware configuration have indicated that the developed algorithm is capable to reliably maintain < 1 ms running time in severe tool-bone collisions, both sampled at 10243 resolutions. The results also showed the algorithm running time has a low sensitivity to bone sampling resolution. The comparisons performed suggested that the proposed approach is significantly faster than comparable methods while relying on lower or similar memory requirements. ConclusionsThe algorithm proposed through this study enables a higher numerical efficiency and is capable to significantly enlarge the maximum resolution that can be used by high fidelity/high realism haptic simulators targeting surgical orthopaedic procedures.

Clozapine-associated secondary antibody deficiency.

Clozapine has recently been described as a novel cause of secondary antibody deficiency (SAD), associated with long-term therapy. Here we critically review the evidence linking clozapine use to an increased infection risk, describe immunological alterations, and discuss potential mechanisms. Individuals with schizophrenia are at two to five times more likely to develop pneumonia than the general population, in particular, when receiving clozapine. Delayed-onset distinguishes clozapine-associated hypogammaglobulinaemia from agranulocytosis or neutropenia that occur at lesser frequency. Biomarker searches in treatment-resistant schizophrenia highlight an immune signature associated with long-term clozapine use. This includes reduction in class-switched memory B cells, echoing common variable immunodeficiency. Recent identification of a role for dopamine in T follicular helper-B cell interactions may inform future clinical studies. The detrimental impact of the increased infection risk associated with clozapine necessitates a re-evaluation of the current monitoring strategies as well as further studies to better understand the underlying mechanisms of SAD in this setting. On the basis of available evidence, we suggest simple modifications to clozapine monitoring including integration of routine vaccination, smoking cessation, and assessment of humoral immunity. Further studies are required to understand the role of clozapine in neuroinflammation as well as other potentially autoantibody-mediated diseases.

Sparsity-Aware Precorrected Tensor Train Algorithm for Fast Solution of 2-D Scattering Problems and Current Flow Modeling on Unstructured Meshes

Acceleration of the method of moments (MoM) solution of the volume integral equation (VIE) on unstructured meshes is performed using a precorrected tensor train (P-TT) algorithm. The elements of the MoM’s unstructured mesh are projected onto a regular Cartesian grid. This enables representation of the MoM matrix as the Toeplitz matrix of point-to-point interactions pre- and post-multiplied by sparse matrices projecting MoM’s basis and testing functions on the Cartesian grid. The Toeplitz matrix is subsequently cast into the form of a multidimensional tensor. The latter is decomposed into the product of smaller dimensional matrices also known as tensor train (TT). TT allows to store Toeplitz matrix in $O(\log N)$ memory for VIE with the Laplace kernel and in $O(N \log N)$ memory for VIE with the Helmholtz kernel. Unlike the FFT-based fast algorithms, the P-TT method enables further memory reduction due to the sparsity of sources on the Cartesian grid. This sparsity pattern can be accounted for in construction of TT cores. It further reduces memory use as well as CPU time required for the multiplication of the Toeplitz matrix with a vector. It is shown that upon sufficient sparsity in representation of the sources on the grid, the P-TT algorithm can outperform both in CPU time and memory its FFT-based counterpart known as the precorrected FFT algorithm.

Hardware Design of a Context-Preserving Filter-Reorganized CNN for Super-Resolution

This paper presents a hardware design of a CNN for single image super-resolution (SISR) processing. Very deep convolutional network for image super-resolution (VDSR) is a promising algorithm for SISR generating high-quality images but it is too complex to be implemented in hardware for commercial products. The paper presents a novel hardware design that is applicable to commercial products with low hardware resource usage while avoiding a significant degradation of image quality. The proposed design employs sub-pixel convolution layers to support various scale factors. A context-preserving 1D reorganization scheme is proposed to reduce the number of multipliers. Additional optimization technique is proposed to reduce the internal SRAM to store the input feature map. Thanks to the proposed optimizations along with additional techniques popular in hardware design, the number of multipliers and the size of SRAMs are reduced to 1/118.27 and 1/47.59, respectively, in comparison with a straightforward implementation of VDSR. The proposed CNN hardware synthesized with TSMC 65-nm CMOS technology, and it requires 8153.1K gate counts and 333.1kB SRAM to process a full-HD image. The proposed context-preserving 1D reorganization and memory reduction optimizations are also applicable to denoising and image classification without a significant performance degradation.

Disrupters as Well as Monitors: Roles of Others During and After Collaborative Remembering in the DRM Procedure.

Influence of others on true and inauthentic memory both during and after collaborative remembering have drawn extensive attention in recent years. Collaborative research has recorded three typical effects: collaborative inhibition and error pruning (i.e., nominal groups recall more true information but also bear higher erroneous intrusions than collaborative groups) during collaboration, as well as post-collaborative recall benefit after collaboration. This study introduced Deese-Roediger-McDermott (DRM) lists not only to investigate these phenomena in semantically related information, but also set a course to explore false memory in the collaborative context. Another issue is the sensitivity of these effects to different episodic memory tests (i.e., item memory and source memory tests). In views of these, the current study instructed participants to study several DRM lists and then recall previously studied words (item recall) together with their displayed colors (source retrieval) twice (Recall 1 and 2). Recall 1 was performed either individually or collaboratively, whereas Recall 2 was conducted individually. The cost of collaborative inhibition was obtained, along with three different beneficial effects: error pruning, false memory reduction, and post-collaborative recall benefit. Furthermore, the novel implication of the current study is that it reveals the sensitivity of collaborative inhibition and error pruning in DRM lists to testing conditions and demonstrates that the modulation of collaboration on false memory occurs in the same way both during and after collaboration. These results are discussed in terms of the retrieval strategy disruption hypothesis and other accounts.

Time and Area Efficient 2-D DWT using Multiplier-less Canonic Signed Digit Technique

A few models have been recommended for proficient VLSI usage of 2-D DWT for constant applications. It is disc overed that multipliers devour more chip zone and expands multifaceted nature of the DWT design. Multiplier-less equipment usage approach gives an answer for diminish chip region, lower equipment intricacy and higher throughput of calculation of the DWT design.The proposed design outline is (i) priority must be given for memory complexity optimization over the arithmetic complexity optimization or reduction of cycle period and (ii) memory utilization efficiency to be considered ahead of memory reduction due to design complexity of memory optimization method. Based on the proposed design outline four separate design approaches and concurrent architectures are presented in this thesis for area-delay and power efficient realization of multilevel 2-D DWT.In this theory a multiplier-less VLSI engineering is proposed utilizing new dispersed number juggling calculation named CSD. We show that CSD is an effective engineering with adders as the principle part and free of ROM, duplication, and subtraction. The proposed design utilizing CSD gives less postponement and least number of cut thought about the current engineering. The reenactment was performed utilizing XILINX 14.1i and ModelSim test system.