Required Memory Space Research Articles

Reducing neural network complexity via optimization algorithms for fault diagnosis in renewable energy systems

This article discusses the challenges involved in diagnosing faults in renewable energy systems. A significant challenge is the high computational demands of the artificial intelligence algorithms needed for grid-connected photovoltaic (GCPV) and wind energy conversion (WEC) systems in fault diagnosis. To address this issue, several methods are proposed to reduce computation time, minimize memory requirements, and improve efficiency. First, optimization algorithms such as the Salp Swarm Algorithm (SSA), Genetic Algorithm (GA), and Particle Swarm Optimization (PSO), combined with machine learning classifiers for feature selection, are suggested to reduce computational time and memory space requirements. This approach notably decreases computation time but has a limited impact on memory space. Secondly, further reduction in memory space requirements is recommended by using the variogram method for data reduction. This method can leverage the outputs of preceding algorithms to optimize renewable energy systems, making their operations cost-effective and efficient. Finally, the outputs of the variogram are used to train neural network (NN), recurrent neural network (RNN), and long short-term memory (LSTM) classifiers to differentiate between various modes of operation in GCPV and WEC systems. Experimental results demonstrate the effectiveness and robustness of the proposed methods, showing that memory space can be reduced by 1.3 to 5.6 times and CPU time by 1.1 to 11 times while maintaining accuracy and improving efficiency compared to conventional algorithms.

Compression of Convolutional Neural Networks With Divergent Representation of Filters.

Convolutional neural networks (CNNs) have made remarkable achievements in many tasks. However, most of them are hardly applied to embedded systems directly because of the requirement of huge memory space and computing power. In this article, we propose a pruning framework, namely, FiltDivNet, to accelerate and compress CNN models for their applicability to small or portable devices. The correlations among filters are taken into account and measured by the goodness of fit. On this basis, a hybrid-cluster pruning strategy is designed with dynamic pruning ratios for different clusters in CNN models. It aims at representing its filters in their diversity by removing redundant ones cluster by cluster. In addition, a new loss function with adaptive sparsity constraints is introduced for the retraining and fine-tuning in the FiltDivNet. Finally, some comparative experiments based on classical CNN models are carried out to demonstrate its effectiveness in compression performance and its adaptability with different CNN architectures.

Part2Point: A Part-Oriented Point Cloud Reconstruction Framework.

Three-dimensional object modeling is necessary for developing virtual and augmented reality applications. Traditionally, application engineers must manually use art software to edit object shapes or exploit LIDAR to scan physical objects for constructing 3D models. This is very time-consuming and costly work. Fortunately, GPU recently provided a cost-effective solution for massive data computation. With GPU support, many studies have proposed 3D model generators based on different learning architectures, which can automatically convert 2D object pictures into 3D object models with good performance. However, as the demand for model resolution increases, the required computing time and memory space increase as significantly as the parameters of the learning architecture, which seriously degrades the efficiency of 3D model construction and the feasibility of resolution improvement. To resolve this problem, this paper proposes a part-oriented point cloud reconstruction framework called Part2Point. This framework segments the object's parts, reconstructs the point cloud for individual object parts, and combines the part point clouds into the complete object point cloud. Therefore, it can reduce the number of learning network parameters at the exact resolution, effectively minimizing the calculation time cost and the required memory space. Moreover, it can improve the resolution of the reconstructed point cloud so that the reconstructed model can present more details of object parts.

BISDU: A Bit-Serial Dot-Product Unit for Microcontrollers

Low-precision quantized neural networks (QNNs) reduce the required memory space, bandwidth, and computational power, and hence are suitable for deployment in applications such as IoT edge devices. Mixed-precision QNNs, where weights commonly have lower precision than activations or different precision is used for different layers, can limit the accuracy loss caused by low-bit quantization, while still benefiting from reduced memory footprint and faster execution. Previous multiple-precision functional units supporting 8-bit, 4-bit, and 2-bit SIMD instructions have limitations, such as large area overhead, under-utilization of multipliers, and wasted memory space for low and mixed bit-width operations. This article introduces BISDU, a bit-serial dot-product unit to support and accelerate execution of mixed-precision low-bit QNNs on resource-constrained microcontrollers. BISDU is a multiplier-less dot-product unit, with frugal hardware requirements (a population count unit and 2:1 multiplexers). The proposed bit-serial dot-product unit leverages the conventional logical operations of a microcontroller to perform multiplications, which enables efficient software implementations of binary ( Xnor ), ternary ( Xor ), and mixed-precision [W×A] ( And ) dot-product operations. The experimental results show that BISDU achieves competitive performance compared to two state-of-the-art units, XpulpNN and Dustin, when executing low-bit-width CNNs. We demonstrate the advantage that bit-serial execution provides by enabling trading accuracy against weight footprint and execution time. BISDU increases the area of the ALU by 68% and the ALU power consumption by 42% compared to a baseline 32-bit RISC-V (RV32IC) microcontroller core. In comparison, XpulpNN and Dustin increase the area by 6.9× and 11.1× and the power consumption by 3.8× and 5.97×, respectively. The bit-serial state-of-the-art, based on a conventional popcount instruction, increases the area by 42% and power by 32%, with BISDU providing a 37% speedup over it.

Combining fragmentation method and high-performance computing: Geometry optimization and vibrational spectra of proteins.

Exploring the structures and spectral features of proteins with advanced quantum chemical methods is an uphill task. In this work, a fragment-based molecular tailoring approach (MTA) is appraised for the CAM-B3LYP/aug-cc-pVDZ-level geometry optimization and vibrational infrared (IR) spectra calculation of ten real proteins containing up to 407 atoms and 6617 basis functions. The use of MTA and the inherently parallel nature of the fragment calculations enables a rapid and accurate calculation of the IR spectrum. The applicability of MTA to optimize the protein geometry and evaluate its IR spectrum employing a polarizable continuum model with water as a solvent is also showcased. The typical errors in the total energy and IR frequencies computed by MTA vis-à-vis their full calculation (FC) counterparts for the studied protein are 5-10 millihartrees and 5 cm-1, respectively. Moreover, due to the independent execution of the fragments, large-scale parallelization can also be achieved. With increasing size and level of theory, MTA shows an appreciable advantage in computer time as well as memory and disk space requirement over the corresponding FCs. The present study suggests that the geometry optimization and IR computations on the biomolecules containing ∼1000 atoms and/or ∼15 000 basis functions using MTA and HPC facility can be clearly envisioned in the near future.

Dynamic method to optimize memory space requirement in real-time video surveillance using convolution neural network

ABSTRACT Real-time video surveillance is one of the most effective ways to observe crime, mischief, and violence. But, most of the recent surveillance system consumes huge memory space to store the video. This article proposed an advanced dynamic video surveillance strategy to utilize minimum memory space with the finest detection of suspicious moving objects. In the proposed system, a convolution neural network (CNN) classifier and frame resolution switching is used to detect the movement of objects with appropriate frame resolution. A high-definition (HD) frame can be recorded dynamically when any movement of a suspicious object is detected. The system records low-quality frames, which are less important. Finally, the improved gradient-based histogram equalization technique is applied to all frames to obtain enhanced suspicious frames. Several real-time imperial tests are conducted and it is observed that the proposed system detects suspicious objects with 98.25% accuracy. Besides, the system consumes 80% less memory storage.

A new method based on swarm intelligence with encrypted data in wireless sensor networks

Wireless sensor networks (WSNs) technology is one of the most essential Internet of things technologies. It is utilized efficiently in a variety of real-world applications, including healthcare, environmental monitoring, tracking, etc. WSNs are composed of sensor nodes with restricted resources. However, the communication between WSN components is not secure. Therefore, it is necessary to build efficient and lightweight cryptographic algorithms to secure shared data. Our paper comprises proposes a secure protocol called grasshopper optimization algorithm routing protocol (GOARP) with a lightweight encryption method in each sensor called rivest cipher 5 (RC5) to enhance network efficiency and simulation in terms of power consumption, required memory space, and computational time. Subsequently, the network lifetime result achieved in the proposed method is about (70%) more than in GOA elliptic curve cryptographic and Diffie Hellman (GOA-ECCDH).

Tacchi: A Pluggable and Low Computational Cost Elastomer Deformation Simulator for Optical Tactile Sensors

Simulation is widely applied in robotics research to save time and resources. There have been several works to simulate optical tactile sensors that leverage either a smoothing method or Finite Element Method (FEM). However, elastomer deformation physics is not considered in the former method, whereas the latter requires a massive amount of computational resources like a computer cluster. In this work, we propose a pluggable and low computational cost simulator using the Taichi programming language for simulating optical tactile sensors, named as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Tacchi</i> . It reconstructs elastomer deformation using particles, which allows deformed elastomer surfaces to be rendered into tactile images and reveals contact information without suffering from high computational costs. Tacchi facilitates creating realistic tactile images in simulation, e.g., ones that capture wear-and-tear defects on object surfaces. In addition, the proposed Tacchi can be integrated with robotics simulators for a robot system simulation. Experiment results showed that Tacchi can produce images with better similarity to real images and achieved higher Sim2Real accuracy compared to the existing methods. Moreover, it can be connected with MuJoCo and Gazebo with only the requirement of 1G memory space in GPU compared to a computer cluster applied for FEM. With Tacchi, physical robot simulation with optical tactile sensors becomes possible.

Methods of Model Order Reduction for Coupled Systems Applied to a Brake Disc‐Wheel Composite

AbstractIn this contribution, investigations on model order reduction for coupled systems composed from components of a passenger car are shown. In today's development processes, the simulation of mechanical components is indispensable and large Finite Element models are often used for this purpose. For the calculation of time‐domain or frequency‐domain analyses, for example, a lot of computing power is required. However, with the application of model order reduction methods, this effort can be reduced, but this results in a trade‐off between the reduction error and the computational time. Since the computation of reduction bases for complete systems can be computationally expensive, it is of interest to be able to reduce components individually and then assemble them into a reduced overall model. This can result in both, a saving of computational effort when creating the bases, as well as a saving of the required memory space. Furthermore, there are many possible combinations of components in the modular systems of today's automotive industry, which emphasizes the model order reduction by parts and not by assemblies. In this work, methods of model order reduction for coupled systems are presented and will be tested on components in the chassis of a sports car. Therefore, an assembly consisting of a brake disc and wheel rim together with the wheel hub are investigated. For this purpose, the eigenmodes and transfer functions of the overall model, the reduced overall model and the assembly built from individual reduced bodies are compared.

Multi-task learning on the edge for effective gender, age, ethnicity and emotion recognition

More and more real-world applications, especially in cognitive robotics, require the running of different computer vision algorithms in parallel on board of embedded devices with limited GPU and memory resources. Multi-task learning, namely the usage of the same model to perform multiple classification and/or regression tasks by learning a shared low level representation, revealed to be a valid solution to reduce computation and required memory space while preserving the accuracy. In this paper, we propose a solution for real-time user profiling based on a multi-task convolutional neural network (CNN) for gender, age, ethnicity and emotion recognition from face images. To find the best trade-off between accuracy and processing time, we evaluate three different architectures, specifically designed for the purpose, and backbones, based on MobileNet, ResNet and SENet, which include convolutional layers, residual blocks and attention modules that already demonstrated great potential in face analysis. We trained the multi-task neural network with a custom learning procedure, which solves the problems of missing labels, dataset imbalance and loss function imbalance through label masking, batch balancing and a custom weighted loss function; there are no other multi-task neural networks for face analysis that address all these challenges simultaneously.. The proposed solution demonstrated its effectiveness in the comparison with the corresponding single-task CNNs in terms of accuracy, processing time and memory space; in fact, the multi-task CNNs achieved a processing speed-up between 2.5 and 4 times and a reduction of the memory space between 2 and 4 times, while preserving the accuracy. Moreover, the useful insights that arise from the experiments allow to choose a solution for face analysis easily integrable into real applications on smart cameras and embedded systems and most suited for the specific application constraints in terms of computational resources.

A linear algorithm for semi-external cutnode computation

In the literature, many algorithms have been proposed for finding cutnodes on undirected graphs, since cutnodes are crucial to graph connectivity. Here, a cutnode of an undirected graph G is a node of G, whose deletion will cause a reachable pair of the other nodes in G to be unreachable. Currently, the difficulty of maintaining the entire G in the main memory makes researchers pay attention to compute cutnodes on semi-external memory model. This paper shows that traditional semi-external algorithms are limited by their unbounded time and I/O consumption, making them impractical when G is relatively large or complex. Thus, we propose a linear semi-external cutnode computation algorithm, named SECN. Assuming that G has n nodes and m edges, and B is the disk block size. SECN is the first that can find all the cutnodes of G in O(m+n) time and with O(mB) I/O cost on the semi-external memory model, as far as we know. SECN also has a smaller minimum memory space requirement than traditional algorithms. Our experimental evaluation conducted on both synthetic and real graphs confirms that SECN significantly outperforms existing algorithms (up to 103 times faster).

PipeNTT: A Pipelined Number Theoretic Transform Architecture

Polynomial multiplication is the key and time-consuming operation among various operators in Post-Quantum Cryptography (PQC), which aims to find quantum-resistant algorithms to prevent attacks launched by quantum computers. Number Theoretic Transform (NTT) is an efficient algorithm that can accelerate the polynomial multiplication from <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {O}(n^{2})$ </tex-math></inline-formula> to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {O}(nlog(n))$ </tex-math></inline-formula> . In this brief, we present a pipelined NTT (PipeNTT) hardware architecture in FPGA to achieve high throughput with fewer hardware resources. The dataflow and the butterfly unit are optimized to minimize the latency. To fulfil the proposed dataflow, a Block RAM (BRAM) based reordering unit is designed to further reduce the hardware resource. Moreover, our architecture can also be applied to Inverse-NTT (INTT). Compared to state-of-the-art parallel designs, our design achieves a 30% lower area-time product with 3x less memory space requirement.

Construction of Computer Algorithms in Bioinformatics of the Fusion Genetic Algorithm

With the continuous in-depth exploration of life sciences, bioinformatics based on life sciences, computer algorithms, and statistics have gradually developed. The research of bioinformatics mainly focuses on the study of genes, and the structural characteristics of genes lead to a large amount of extremely complex data in the study of bioinformatics. Analyzing data in bioinformatics research requires accurate calculation by computer algorithms. However, common computer algorithms such as the dynamic programming algorithm and the genetic algorithm have the disadvantages of large memory or inaccurate optimization. Combining the ant colony algorithm (ACA) and GA can give the advantages of the two methods that should be fully utilized to efficiently analyze the biological information data. In this paper, the ant colony fusion genetic algorithm (ACA-GA), GA, and dynamic programming algorithm are used to compare and analyze the sensitivity, convergence speed, sequence alignment accuracy, and required memory space of gene sequences. The experimental results show that compared with ACA-GA, the dynamic programming algorithm has the advantage of finding the optimal alignment of 100%, but the memory required is too large, the memory required is more than 10 times that of the ant colony fusion genetic algorithm and the sensitivity is not as good as ACA-GA. The convergence speed of ACA-GA is faster than the gene comparison speed of GA and the accuracy is 2.6% better than that of GA on average. ACA-GA has the advantages of GA and ACA, which can improve the computational efficiency of biological data in bioinformatics.

A parallel computing method for the higher order tensor renormalization group

In this paper, we propose a parallel computing method for the Higher Order Tensor Renormalization Group (HOTRG) applied to a d-dimensional (d≥2) simple lattice model. Sequential computation of the HOTRG requires O(χ4d−1) computational cost, where χ is bond dimension, in a step to contract indices of tensors. When we simply distribute elements of a local tensor to each process in parallel computing of the HOTRG, frequent communication between processes occurs. The simplest way to avoid such communication is to hold all the tensor elements in each process, however, it requires O(χ2d) memory space. In the presented method, placement of a local tensor element to more than one process is accepted and sufficient local tensor elements are distributed to each process to avoid communication between processes during considering computation step. For the bottleneck part of computational cost, such distribution is achieved by distributing elements of two local tensors to χ2 processes according to one of the indices of each local tensor which are not contracted during considering computation. In the case of d≥3, computational cost in each process is reduced to O(χ4d−3) and memory space requirement in each process is kept to be O(χ2d−1).

Projection-to-image transform frame: a lightweight block reconstruction network for computed tomography

Several reconstruction networks have been invented to solve the problem of learning-based computed tomography (CT) reconstruction. However, the application of neural networks to tomographic reconstruction remains challenging due to unacceptable memory space requirements. In this study, we present a novel lightweight block reconstruction network (LBRN), which transforms the reconstruction operator into a deep neural network by unrolling the filter back-projection (FBP) method. Specifically, the proposed network contains two main modules, which respectively correspond to the filter and back-projection of the FBP method. The first module of the LBRN decouples the relationship of the Radon transform between the reconstructed image and the projection data. Therefore, the following module, block back-projection, can use the block reconstruction strategy. Because each image block is only connected with part-filtered projection data, the network structure is greatly simplified and the parameters of the whole network are dramatically reduced. Moreover, this approach is trained end-to-end, working directly from raw projection data, and does not depend on any initial images. Five reconstruction experiments are conducted to evaluate the performance of the proposed LBRN: full angle, low-dose CT, region of interest, metal artifact reduction and a real data experiment. The results of the experiments show that the LBRN can be effectively introduced into the reconstruction process and has outstanding advantages in terms of different reconstruction problems.

Power system transient security assessment based on deep learning considering partial observability

Due to the ever-increasing complexity of power systems, transient stability analysis has been reshaped into a complicated and critical issue. The real-time identification of the transient stability status would assist the power system operators in decision-making during critical events. This paper proposes a combined transient security assessment (TSA) method based on convolutional neural networks (CNNs) and a novel feature selection technique using sensitivity analysis. CNN benefits from the keen ability to understand raw data features, which decreases the required memory space. The proposed sensitivity analysis technique for feature selection can lead to a trade-off between dependency on measurable data, accuracy, and computational complexity. Furthermore, in the proposed method, an information-theoretic measure, namely mutual information (MI), is used to handle partial observability conditions. The simulation results validate the proposed method's accuracy and robustness using several metrics and comparisons with the state-of-the-art and previously presented methods.

An analysis of heuristic metrics for classifier ensemble pruning based on ordered aggregation

Classifier ensemble pruning is a strategy through which a subensemble can be identified via optimizing a predefined performance criterion. Choosing the optimum or suboptimum subensemble decreases the initial ensemble size and increases its predictive performance. In this article, a set of heuristic metrics will be analyzed to guide the pruning process. The analyzed metrics are based on modifying the order of the classifiers in the bagging algorithm, with selecting the first set in the queue. Some of these criteria include general accuracy, the complementarity of decisions, ensemble diversity, the margin of samples, minimum redundancy, discriminant classifiers, and margin hybrid diversity. The efficacy of those metrics is affected by the original ensemble size, the required subensemble size, the kind of individual classifiers, and the number of classes. While the efficiency is measured in terms of the computational cost and the memory space requirements. The performance of those metrics is assessed over fifteen binary and fifteen multiclass benchmark classification tasks, respectively. In addition, the behavior of those metrics against randomness is measured in terms of the distribution of their accuracy around the median. Results show that ordered aggregation is an efficient strategy to generate subensembles that improve both predictive performance as well as computational and memory complexities of the whole bagging ensemble.

IFDDA, an easy-to-use code for simulating the field scattered by 3D inhomogeneous objects in a stratified medium: tutorial

In this tutorial, we introduce a solver of monochromatic Maxwell equations made freely available at https://www.fresnel.fr/perso/chaumet/ifdda.html, based on the volume moment method. The Institut Fresnel Discrete Dipole Approximation or Idiot-Friendly Discrete Dipole Approximation (IFDDA) calculates the diffracted field, the optical forces, and the image through a microscope of any three-dimensional inhomogeneous object, possibly anisotropic, placed in a stratified medium. In this method, only the object is meshed so the required memory space is kept to a minimum. We describe the principle and the potentialities of IFDDA and present comparisons with Mie theory and experimental data to assess the accuracy of the method. In addition, we provide a user guide for first steps with the solver. We hope that you will use and enjoy this numerical tool!

Acceleration of reconstruction for compressed sensing based synthetic transmit aperture imaging by using in-phase/quadrature data

Compressed sensing-based synthetic transmit aperture (CS-STA) was previously proposed to recover the full radio-frequency (RF) channel dataset of synthetic transmit aperture (STA) from that of a smaller number of randomly apodized plane wave (PW) transmissions. In this way, the imaging frame rate (FR) and contrast are improved with maintained spatial resolution, compared with those of STA. Because CS-STA reconstruction is repeated for all receive elements and RF samples (with a high sampling frequency), the recovery of STA dataset in RF domain is time-consuming. In the meantime, a large amount of RF data needs to be transferred and stored, resulting in an increase of system complexity and required memory space. In this study, CS-STA is extended to in-phase/quadrature (IQ) domain (with lower sampling frequency) for the recovery of baseband STA IQ dataset to accelerate the CS-STA reconstruction by reducing the amount of data to be processed. More importantly, CS-STA reconstruction using IQ data is of practical importance, as clinical ultrasound systems typically record baseband IQ signal instead of RF signal. Simulations, phantom and in vivo experiments verify the feasibility of CS-STA in IQ domain for the recovery of STA dataset. More specifically, CS-STA using IQ data achieves similar image quality and appreciably improves reconstruction speed (by ∼3 times) compared with that using RF data. These findings demonstrate that IQ-domain CS-STA is capable of relieving the computational and storage burdens, which may facilitate the implementation of CS-STA in practical ultrasound systems.

Light-weight network for real-time adaptive stereo depth estimation

Self-supervised learning methods have been proved effective in the task of real-time stereo depth estimation with the requirement of lower memory space and less computational cost. In this paper, a light-weight adaptive network (LWANet) is proposed by combining the self-supervised learning method to perform online adaptive stereo depth estimation for low computation cost and low GPU memory space. Instead of a regular 3D convolution, the pseudo 3D convolution is employed in the proposed light-weight network to aggregate the cost volume for achieving a better balance between the accuracy and the computational cost. Moreover, based on U-Net architecture, the downsample feature extractor is combined with a refined convolutional spatial propagation network (CSPN) to further refine the estimation accuracy with little memory space and computational cost. Extensive experiments demonstrate that the proposed LWANet effectively alleviates the domain shift problem by online updating the neural network, which is suitable for embedded devices such as NVIDIA Jetson TX2. The relevant codes are available at https://github.com/GANWANSHUI/LWANet