Limited Computational Resources Research Articles

HIt Discovery using docking ENriched by GEnerative Modeling (HIDDEN GEM): A novel computational workflow for accelerated virtual screening of ultra-large chemical libraries.

Recent rapid expansion of make-on-demand, purchasable, chemical libraries comprising dozens of billions or even trillions of molecules has challenged the efficient application of traditional structure-based virtual screening methods that rely on molecular docking. We present a novel computational methodology termed HIDDEN GEM (HIt Discovery using Docking ENriched by GEnerative Modeling) that greatly accelerates virtual screening. This workflow uniquely integrates machine learning, generative chemistry, massive chemical similarity searching and molecular docking of small, selected libraries in the beginning and the end of the workflow. For each target, HIDDEN GEM nominates a small number of top-scoring virtual hits prioritized from ultra-large chemical libraries. We have benchmarked HIDDEN GEM by conducting virtual screening campaigns for 16 diverse protein targets using Enamine REAL Space library comprising 37 billion molecules. We show that HIDDEN GEM yields the highest enrichment factors as compared to state of the art accelerated virtual screening methods, while requiring the least computational resources. HIDDEN GEM can be executed with any docking software and employed by users with limited computational resources.

RepDNet: A re-parameterization despeckling network for autonomous underwater side-scan sonar imaging with prior-knowledge customized convolution

Side-scan sonar (SSS) is now a prevalent instrument for large-scale seafloor topography measurements, deployable on an autonomous underwater vehicle (AUV) to execute fully automated underwater acoustic scanning imaging along a predetermined trajectory. However, SSS images often suffer from speckle noise caused by mutual interference between echoes, and limited AUV computational resources further hinder noise suppression. Existing approaches for SSS image processing and speckle noise reduction rely heavily on complex network structures and fail to combine the benefits of deep learning and domain knowledge. To address the problem, RepDNet, a novel and effective despeckling convolutional neural network is proposed. RepDNet introduces two re-parameterized blocks: the Pixel Smoothing Block (PSB) and Edge Enhancement Block (EEB), preserving edge information while attenuating speckle noise. During training, PSB and EEB manifest as double-layered multi-branch structures, integrating first-order and second-order derivatives and smoothing functions. During inference, the branches are re-parameterized into a 3 × 3 convolution, enabling efficient inference without sacrificing accuracy. RepDNet comprises three computational operations: 3 × 3 convolution, element-wise summation and Rectified Linear Unit activation. Evaluations on benchmark datasets, a real SSS dataset and Data collected at Lake Mulan aestablish RepDNet as a well-balanced network, meeting the AUV computational constraints in terms of performance and latency.

AdaStreamLite

Streaming speech recognition aims to transcribe speech to text in a streaming manner, providing real-time speech interaction for smartphone users. However, it is not trivial to develop a high-performance streaming speech recognition system purely running on mobile platforms, due to the complex real-world acoustic environments and the limited computational resources of smartphones. Most existing solutions lack the generalization to unseen environments and have difficulty to work with streaming speech. In this paper, we design AdaStreamLite, an environment-adaptive streaming speech recognition tool for smartphones. AdaStreamLite interacts with its surroundings to capture the characteristics of the current acoustic environment to improve the robustness against ambient noise in a lightweight manner. We design an environment representation extractor to model acoustic environments with compact feature vectors, and construct a representation lookup table to improve the generalization of AdaStreamLite to unseen environments. We train our system using large speech datasets publicly available covering different languages. We conduct experiments in a large range of real acoustic environments with different smartphones. The results show that AdaStreamLite outperforms the state-of-the-art methods in terms of recognition accuracy, computational resource consumption and robustness against unseen environments.

Enhancing trust and security in IoT computing offloading through game theory and blockchain‐based control strategy

SummaryCollaboration between IoT end‐devices and edge servers or idle devices enables offloading of computational tasks, providing an effective solution to address inherent limitations in computational resources, storage capacity, and energy efficiency. However, IoT network openness introduces security challenges, such as privacy breaches, data security, and “free‐riding” attack. Securing computational offloading is crucial for service quality and reliability. In this article, we introduce a comprehensive approach to address these challenges. First, trust metrics are performed at the level of resource requester and resource provider respectively, and the trust level of both parties is matched to reduce the adverse effects of malicious devices or servers on computing offloading. Then, a distributed decision‐making method based on game theory is designed to improve the performance of the algorithm and avoid the single‐point‐of‐failure problem through the joint participation of multiple edge servers in decision‐making. Finally, a blockchain‐based task offloading process control strategy is used to deal with the “free‐riding” attack. The results of simulation experiments show the effectiveness of our proposed offloading decision‐making method.

Real-Time Object Detection and Tracking for Unmanned Aerial Vehicles Based on Convolutional Neural Networks

This paper presents a system applied to unmanned aerial vehicles based on Robot Operating Systems (ROSs). The study addresses the challenges of efficient object detection and real-time target tracking for unmanned aerial vehicles. The system utilizes a pruned YOLOv4 architecture for fast object detection and the SiamMask model for continuous target tracking. A Proportional Integral Derivative (PID) module adjusts the flight attitude, enabling stable target tracking automatically in indoor and outdoor environments. The contributions of this work include exploring the feasibility of pruning existing models systematically to construct a real-time detection and tracking system for drone control with very limited computational resources. Experiments validate the system’s feasibility, demonstrating efficient object detection, accurate target tracking, and effective attitude control. This ROS-based system contributes to advancing UAV technology in real-world environments.

Driver Distraction Detection Based on Cloud Computing Architecture and Lightweight Neural Network

Distracted behavior detection is an important task in computer-assisted driving. Although deep learning has made significant progress in this area, it is still difficult to meet the requirements of the real-time analysis and processing of massive data by relying solely on local computing power. To overcome these problems, this paper proposes a driving distraction detection method based on cloud–fog computing architecture, which introduces scalable modules and a model-driven optimization based on greedy pruning. Specifically, the proposed method makes full use of cloud–fog computing to process complex driving scene data, solves the problem of local computing resource limitations, and achieves the goal of detecting distracted driving behavior in real time. In terms of feature extraction, scalable modules are used to adapt to different levels of feature extraction to effectively capture the diversity of driving behaviors. Additionally, in order to improve the performance of the model, a model-driven optimization method based on greedy pruning is introduced to optimize the model structure to obtain a lighter and more efficient model. Through verification experiments on multiple driving scene datasets such as LDDB and Statefarm, the effectiveness of the proposed driving distraction detection method is proved.

Whole genome sequencing in the palm of your hand: how to implement a MinION Galaxy-based workflow in a food safety laboratory for rapid Salmonella spp. serotyping, virulence, and antimicrobial resistance gene identification.

Whole Genome Sequencing (WGS) implementation in food safety laboratories is a significant advancement in food pathogen control and outbreak tracking. However, the initial investment for acquiring next-generation sequencing platforms and the need for bioinformatic skills represented an obstacle for the widespread use of WGS. Long-reading technologies, such as the one developed by Oxford Nanopore Technologies, can be easily implemented with a minor initial investment and with simple protocols that can be performed with basic laboratory equipment. Herein, we report a simple MinION Galaxy-based workflow with analysis parameters that allow its implementation in food safety laboratories with limited computer resources and without previous knowledge in bioinformatics for rapid Salmonella serotyping, virulence, and identification of antimicrobial resistance genes. For that purpose, the single use Flongle flow cells, along with the MinION Mk1B for WGS, and the community-driven web-based analysis platform Galaxy for bioinformatic analysis was used. Three strains belonging to three different serotypes, monophasic S. Typhimurium, S. Grancanaria, and S. Senftenberg, were sequenced. After 24 h of sequencing, enough coverage was achieved in order to perform de novo assembly in all three strains. After evaluating different tools, Flye de novo assemblies with medaka polishing were shown to be optimal for in silico Salmonella spp. serotyping with SISRT tool followed by antimicrobial and virulence gene identification with ABRicate. The implementation of the present workflow in food safety laboratories with limited computer resources allows a rapid characterization of Salmonella spp. isolates.

Model Pruning Enables Efficient Federated Learning on Edge Devices.

Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving data privacy, which has wide applicability to image and vision applications. A challenge is that client devices in FL usually have much more limited computation and communication resources compared to servers in a data center. To overcome this challenge, we propose PruneFL -a novel FL approach with adaptive and distributed parameter pruning, which adapts the model size during FL to reduce both communication and computation overhead and minimize the overall training time, while maintaining a similar accuracy as the original model. PruneFL includes initial pruning at a selected client and further pruning as part of the FL process. The model size is adapted during this process, which includes maximizing the approximate empirical risk reduction divided by the time of one FL round. Our experiments with various datasets on edge devices (e.g., Raspberry Pi) show that: 1) we significantly reduce the training time compared to conventional FL and various other pruning-based methods and 2) the pruned model with automatically determined size converges to an accuracy that is very similar to the original model, and it is also a lottery ticket of the original model.

Communication and Energy Efficient Decentralized Learning Over D2D Networks

Device-to-device (D2D)-assisted decentralized learning has been proposed for mobile devices to collaboratively train artificial intelligence networks without the centralized parameter server. However, a densely connected network will cause large learning latency and energy consumption due to the limited computation and communication resources. In addition, link selection and aggregation weight have a significant impact on the learning performance. To cope with these challenges, we propose a joint computing power adjustment, wireless resource allocation, link selection, and aggregation weight adaptation mechanism to improve both communication and energy efficiencies. Specifically, the learning performances including the convergence rate, per-iteration learning latency, and per-iteration energy consumption are first analyzed. Then, an optimization problem is formulated to minimize the total learning cost, which is defined as the weighted sum of total learning latency and energy consumption. Given a network topology, the computing power and wireless resource allocation are optimized by the alternating optimization algorithm. Moreover, the optimal aggregation weight is obtained by semidefinite programming. With respect to link selection, we propose a tabu search based meta-heuristic algorithm to approximately achieve feasible solutions with a low computational complexity. Finally, extensive experiments demonstrate that the proposed link selection algorithm can significantly reduce the learning cost under the given learning accuracy requirement.

A module placement scheme for fog-based smart farming applications

<span lang="EN-US">As in Industry 4.0 era, the impact of the internet of things (IoT) on the advancement of the agricultural sector is constantly increasing. IoT enables automation, precision, and efficiency in traditional farming methods, opening up new possibilities for agricultural advancement. Furthermore, many IoT-based smart farming systems are designed based on fog and edge architecture. Fog computing provides computing, storage, and networking services to latency-sensitive applications (such as Agribots-agricultural robots-drones, and IoT-based healthcare monitoring systems), instead of sending data to the cloud. However, due to the limited computing and storage resources of fog nodes used in smart farming, designing a modules placement scheme for resources management is a major challenge for fog based smart farming applications. In this paper, our proposed module placement algorithm aims to achieve efficient resource utilization of fog nodes and reduce application <span>delay and network usage in Fog-based smart farming applications. To evaluate the efficacy of our proposal, the simulation was done using iFogSim. Results show that the proposed approach is able to achieve significant</span> reductions in latency and network usage.</span>

QANS: Toward Quantized Neural Network Adversarial Noise Suppression

Neural network quantization techniques play an important role in efficiently deploying deep learning models on the hardware with limited computing and storage resources. Numerous applications of this technology, such as autopilot, necessitate not just efficiency, but also robustness. Research on the robustness of quantized networks against adversarial attacks is becoming one of the major points of interest. In this work, we rethink the impact of quantization on adversarial attacks and explore the boundary of the robustness of quantized neural networks. This study reveals that activation quantization can be used as a defense to weaken adversarial noise, but the robustness of quantized models is still limited by the amplification effect of network errors, including quantization errors and adversarial noise. To address this problem, we propose the Quantization Adversarial Noise Suppression (QANS) method that employs a Gaussian kernel regularization constraint to stabilize the model by restricting the perturbation error within two levels of tolerance. Extensive experiments are conducted with Wide ResNet and VGG-16 models on CIFAR-10 and SVHN datasets under different attack methods, including several white-box and black-box attacks. Experimental results show the proposed method achieves superior robustness to prior works.

CTFCD: Channel transformer based on full convolutional decoder for single image deraining

Although convolutional neural network and visual transformer have been successfully applied in various field of computer vision, there is little work combining them to construct an efficient network model to solve image deraining tasks. Convolutional neural network is utilized to extract the features from each region, while visual transformer is utilized to extract the context information between local features. Due to the limitations in computational resources and processing time, it is difficult for visual transformer to process high-resolution images, which hinders the application of visual transformer in devices with limited hardware resources. The purpose of this article is to utilize the advantages of to design a lightweight encoder-decoder network for real-time image deraining. Firstly, a novel channel Transformer module is designed to obtain global contextual information, where deep separable convolution is utilized to extract multi-scale local features and a Transformer encoder is constructed by stacking Transformer modules. Secondly, a decoder based on a fully convolution is designed to adopt mask attention and inverted bottleneck convolution to achieve progressive feature fusion and feature reconstruction, which significantly reduces computational complexity and memory requirement. A large number of experimental results have verified that the proposed method has superior performance compared with other state-of-the art methods, while the computational cost and parameter quantity are much smaller than those of similar methods.

Sustainable and lightweight domain-based intrusion detection system for in-vehicle network

Intelligent transportation systems are designed to enhance and optimize the traffic flow, safety of urban mobility, and improve energy efficiency. While advanced vehicles are equipped with new features, such as communication technologies for the exchange of safety messages, the communication interfaces of the vehicles increase the attack surfaces for attackers to exploit, even into the in-vehicle network (IVN). The automotive Ethernet and intrusion detection systems (IDSs) are promising solutions to the security problem inside the IVN. The automotive Ethernet has higher bandwidth capacity at an economical cost and flexibility for expansion than the current solution. The IDSs can protect the IVN from attackers compromising the vehicle. They can be implemented based on machine learning to learn from the normal IVN traffic behavior. However, numerous IDSs that utilize machine learning face hardware limitations when training within the in-vehicle network. Thus, the models have to be trained outside the IVN and then imported into it. Moreover, the IVN messages are unlabeled whether they are normal or under attack. We propose a lightweight unsupervised IDS that enables training in the IVN with limited computation resources. Our IDS models have an impressive parameter reduction of up to 94% compared to existing models. This leads to a remarkable reduction in memory usage of up to 86%, training time slashed by up to 69%, and a remarkable drop in energy consumption by up to 68%. Despite the size reduction, the proposed models remain only slightly less accurate than current solutions by up to 2%.

Improving the model robustness of flood hazard mapping based on hyperparameter optimization of random forest

Traditional machine learning algorithms face challenges in assessing flood susceptibility reliably due to their low robustness and the inherent 'black-box' nature. This paper utilizes five hyperparameter optimization algoirthms (HPO), namely grid search (GS), random search (RS), gauss process (GP), tree-structured parzen estimator (TPE) and simulated annealing (SA), to tune the traditional random forest’s (RF) hyperparameters to improve the robustness of flood hazard mapping (FHM) models at Ningxiang City Hunan Province, China. Additionally, SHapley Additive exPlanations (SHAP) method were used to interpret the decision-mechanisms of these flood hazard models. This study considers 19 pluvial flood influencing factors and 2064 flood locations to create a geospatial database. The performance of each hybrid model was evaluated by area under the receiver operating characteristic (ROC) curve (AUC) and several validation methods. The results demonstrate that the developed hybrid models demonstrated good performance, with RF-TPE achieving the highest AUC (0.9660), followed by RF-GP (0.9648), RF-SA (0.9624), RF-GS (0.9612), RF-RS (0.9600), and RF (0.9539). The RF-TPE model exhibits superior robustness than other models, and the FHM constructed using it is more reliable. HPO is an effective approach to improve the predictive accuracy and robustness of FHM models. When considering limited computational resources, Bayesian optimization (TPE) should be prioritized for optimizing FHM models, followed by metaheuristic algorithms and model-free algorithms. Moreover, the study revealed that distance from river, peak rainfall intensity, continuous rainfall, antecedent effective rainfall, and terrain relief, are the most significant for pluvial FHM modeling in this region.

A Lightweight Neural Network for Spectroscopic Ellipsometry Analysis

AbstractEllipsometry is a widely used technique in thin film characterizations. To extract the optical properties of the films from the measured data, regression data fitting techniques have been developed that iteratively find a set of optical parameters that best fit the observations. However, this iterative process often faces challenges in converging to the correct solution, and it can be time‐consuming. To address these issues, an 8‐bit quantized lightweight method of neural network analysis for spectroscopic ellipsometry are proposed. This method features compact neural network modules to enhance speed and efficiency. The effectiveness of the approach through experimental verification is validated on a diverse range of material films, including metals, semiconductors, and dielectrics. The approach is fully automatic and lightweight, which offers a new perspective on balancing predictive accuracy with limited computational resources. This method holds the potential to achieve automatic, rapid, and high‐throughput optical characterization of films, facilitating real‐time quality monitoring for repeatable high‐precision film manufacturing.

Review article: Towards improved drought prediction in the Mediterranean region – modeling approaches and future directions

Abstract. There is a scientific consensus that the Mediterranean region (MedR) is warming and as the temperature continues to rise, droughts and heat waves are becoming more frequent, severe, and widespread. Given the detrimental effects of droughts, it is crucial to accelerate the development of forecasting and early warning systems to minimize their negative impact. This paper reviews the current state of drought modeling and prediction applied in the MedR, including statistical, dynamical, and hybrid statistical–dynamical models. By considering the multifaceted nature of droughts, the study encompasses meteorological, agricultural, and hydrological drought forms and spans a variety of forecast scales, from weekly to annual timelines. Our objective is to pinpoint the knowledge gaps in literature and to propose potential research trajectories to improve the prediction of droughts in this region. The review finds that while each method has its unique strengths and limitations, hybrid statistical–dynamical models appear to hold the most promising potential for skillful prediction with seasonal to annual lead times. However, the application of these methods is still challenging due to the lack of high-quality observational data and the limited computational resources. Finally, the paper concludes by discussing the importance of using a combination of sophisticated methods such as data assimilation techniques, machine learning models, and copula models and of integrating data from different sources (e.g., remote sensing data, in situ measurements, and reanalysis) to improve the accuracy and efficiency of drought forecasting.

Lightweight Intrusion Detection Method of Vehicle CAN Bus under Computational Resource Constraints

In order to improve the security protection performance of the vehicle Controller Area Network (CAN) bus, the research builds an adaptive lightweight intrusion detection algorithm based on the limited computing and storage resources of the on-board ECU environment and the message cycle characteristics to supervise and detect the vehicle CAN bus intrusion. The results showed that the message cycle-based adaptive intrusion detection algorithm had high accuracy and recall rate, and fast computational search efficiency, with a stable detection time of less than 3 seconds. The intrusion detection capability is continuously optimized as the training time increases, and after stabilization, the resource utilization rate reaches over 95% with a throughput of 100Mb/s. The algorithm has strong protection capabilities. The average vehicle CPU usage of the algorithm is only 4.76%, which is 10.17% lower than the intrusion detection algorithm based on support vector machines. It can effectively prevent interference with the normal operation of the vehicle CAN bus. The algorithm has high detection accuracy for interrupt type attacks, and there are no false positives or missed alarms. For injection type attacks, the probability of missed alarms is less than 1%. The intrusion detection of vehicle CAN bus based on the message cycle characteristics provides technical reference for the safety and stability of the vehicle network, and has important practical value for the intelligent and networked development of the automobile industry.

Federated Recognition Architecture Based on Voting and Feedback Mechanisms for Accuracy Object Classification in Distributed Edge Intelligence Environment

Deep learning has recently become a crucial tool to solve many complex problems and has the potential to revolutionize industries. With the widespread adoption of the Internet of Things, there are now many devices with limited computational resources that are capable of running deep learning models. It has opened up new opportunities to implement deep learning in the edge environment so that decisions can be made locally without sending data to a cloud server for processing. However, because of limited resources, the model’s performance and communication overhead are challenges when deploying learning models to edge devices. The focus of this paper is to investigate and propose a federated recognition architecture for object classification in a distributed edge intelligence environment. Specifically, we build an edge server that includes a voting module and a feedback module to improve the overall accuracy of object classification. The voting module aggregates predictions of multiple edge devices whereas the feedback module sends the voting results to edge devices to adjust the local deep learning model. We build edge devices based on the EdgeX platform which makes it easy to manage data and optimize communication overheads. Because the edge server and edge nodes only exchange prediction results, our proposed architecture ensures security with sensitive data as well as deep learning model architecture. By testing on the image dataset, we evaluate the proposed architecture’s performance and show that it outperforms individual local models in terms of accuracy. Furthermore, our experiments demonstrate that with the feedback mechanism, the deep learning model is constantly updated with new data to maintain accuracy and avoid being outdated. Besides, we prove the real-time processing speed by collecting the delay time of the proposed model. The results show that our proposed architecture has the potential to be deployed in practical applications such as smart cities, and surveillance systems.

Human-Centric Resource Allocation for the Metaverse With Multiaccess Edge Computing

Multi-access edge computing (MEC) is a promising solution to the computation-intensive, low-latency rendering tasks of the metaverse. However, how to optimally allocate limited communication and computation resources at the edge to a large number of users in the metaverse is quite challenging. In this paper, we propose an adaptive edge resource allocation method based on multi-agent soft actor-critic with graph convolutional networks (SAC-GCN). Specifically, SAC-GCN models the multi-user metaverse environment as a graph where each agent is denoted by a node. Each agent learns the interplay between agents by graph convolutional networks with self-attention mechanism to further determine the resource usage for one user in the metaverse. The effectiveness of SAC-GCN is demonstrated through the analysis of user experience, balance of resource allocation, and resource utilization rate by taking a virtual city park metaverse as an example. Experimental results indicate that SAC-GCN outperforms other resource allocation methods in improving overall user experience, balancing resource allocation, and increasing resource utilization rate by at least 27%, 11%, and 8%, respectively.

Efficient building damage assessment from post-disaster aerial video using lightweight deep learning models

ABSTRACTA challenging problem for post-disaster emergency services is how to quickly and precisely acquire building damage with limited time and computation resources. With the development of unmanned aerial vehicle (UAV) technology and convolutional neural networks (CNN), using drone video and CNNs to determine the building damage has become one effective solution. In this research, we propose a stacked lightweight damage assessment model to overcome the current challenge in getting fast and precise post-disaster information based on aerial video datasets and artificial intelligence. Specifically, we developed an instance-level recognition building damage dataset based on aerial video. Then, we proposed a stacked lightweight ShuffleNet architecture that includes location and classification models to assess the building damage state. With regards to the location model, the lightweight network achieves a reduction in the model training time by approximately 37% and in the detection time by 44%, achieving similar location accuracy. As for the classification model, the trained model can achieve classification accuracy of approximately 83% for different damage levels by optimizing ShuffleNet. For post-disaster emergencies with limited time and computation resources, the proposed framework provides a valuable solution to better balance fast and precise building damage assessment.