Low Computational Resources Research Articles

An effective surrogate-assisted rank method for evolutionary neural architecture search

Evolutionary neural architecture search (ENAS) is able to automatically design high-performed architectures of deep neural networks (DNNs) for specific tasks. In recent years, surrogate models have gained significant traction because they can estimate the performance of neural architectures, avoiding excessive computational costs for training. However, most existing surrogate models primarily predict the performance of architectures directly or predict pairwise comparison relationships, which makes it challenging to obtain the rank of a group of architectures when training samples are limited. To address this problem, we propose TCMR-ENAS, an effective triple-competition model-assisted rank method for ENAS. TCMR-ENAS employs a novel triple-competition surrogate model combined with a score-based fitness evaluation method to predict group performance rank. Moreover, a progressive online learning method is proposed to enhance the predictive performance of the triple-competition surrogate model in the framework of modified genetic search. To validate the effectiveness of TCMR-ENAS, we conducted a series of experiments on NAS-Bench-101, NAS-Bench-201, NATS-Bench and NAS-Bench-301, respectively. Experimental results show that TCMR-ENAS can achieve better performance with lower computational resources. The accuracies of searched architectures achieve the best results compared with those of the state-of-the-art methods with limited training samples. In addition, the factors that may influence the effectiveness of TCMR-ENAS are explored in the ablation studies.

Research on Autonomous Navigation of Unmanned Aerial Vehicles Based on Correlation-Based Image Comparison Methods

Relevance. Autonomous navigation of unmanned aerial vehicles (UAVs) is one of the key challenges in the modern aerospace industry. Specifically, for small UAVs, the task of autonomous navigation becomes even more complex due to limitations in computational resources and sensor capabilities. Optimizing navigation methods under such conditions is a pressing issue that requires solutions capable of providing high performance with minimal resource consumption.Objective. Improving the efficiency of autonomous navigation for small UAVs by using correlation methods for image comparison. Achieving this objective is linked to the development and evaluation of algorithms that provide high-speed and accurate navigation with limited computational resources. The work uses methods such as the autocorrelation function, Pearson correlation, the structural similarity index, and a simple neural network for image comparison.Solution. The research showed that the autocorrelation-based approach demonstrates the best performance under low computational resources. It ensures high-speed processing of the entire image and shows optimal detection accuracy. Compared to other methods presented in the study, the autocorrelation approach is capable of working not only with noise in the "reference" map but also of using alternative areas with altered patterns as detected regions.The scientific novelty of the study is determined by the systematic comparison of various methods applied to the task of image comparison for small UAVs with limited computational resources. Unlike well-known works in the field of correlation-extreme systems, this research focuses on the use of a "reference map" and a search area, representing two different images of the same terrain taken from different sources. This is a key difference, as most methods are not highly efficient in processing such images where patterns may differ significantly.The practical significance of the developed algorithm lies in the fact that the proposed autocorrelation-based method can be used by developers of autonomous small UAV control systems to reduce computational load and increase data processing speed.

Deep Autoencoder for Real-Time Single-Channel EEG Cleaning and Its Smartphone Implementation Using TensorFlow Lite With Hardware/Software Acceleration.

To remove signal contamination in electroencephalogram (EEG) traces coming from ocular, motion, and muscular artifacts which degrade signal quality. To do this in real-time, with low computational overhead, on a mobile platform in a channel count independent manner to enable portable Brain-Computer Interface (BCI) applications. We propose a Deep AutoEncoder (DAE) neural network for single-channel EEG artifact removal, and implement it on a smartphone via TensorFlow Lite. Delegate based acceleration is employed to allow real-time, low computational resource operation. Artifact removal performance is quantified by comparing corrupted and ground-truth clean EEG data from public datasets for a range of artifact types. The on-phone computational resources required are also measured when processing pre-saved data. DAE cleaned EEG shows high correlations with ground-truth clean EEG, with average Correlation Coefficients of 0.96, 0.85, 0.70 and 0.79 for clean EEG reconstruction, and EOG, motion, and EMG artifact removal respectively. On-smartphone tests show the model processes a 4 s EEG window within 5 ms, substantially outperforming a comparison FastICA artifact removal algorithm. The proposed DAE model shows effectiveness in single-channel EEG artifact removal. This is the first demonstration of a low-computational resource deep learning model for mobile EEG in smartphones with hardware/software acceleration. This work enables portable BCIs which require low latency real-time artifact removal, and potentially operation with a small number of EEG channels for wearability. It makes use of the artificial intelligence accelerators found in modern smartphones to improve computational performance compared to previous artifact removal approaches.

Multi-Activity Step Counting Algorithm Using Deep Learning Foot Flat Detection with an IMU Inside the Sole of a Shoe.

Step counting devices were previously shown to be efficient in a variety of applications such as athletic training or patient's care programs. Various sensor placements and algorithms were previously experimented, with a best mean absolute percentage error (MAPE) close to 1% in simple mono-activity walking conditions. In this study, an existing running shoe was first instrumented with an inertial measurement unit (IMU) and used in the context of multi-activity trials, at various speeds, and including several transition phases. A total of 21 participants with diverse profiles (gender, age, BMI, activity style) completed the trial. The data recorded was used to develop a step counting algorithm based on a deep learning approach, and further validated against a k-fold cross validation process. The results revealed that the step counts were highly correlated to gyroscopes and accelerometers norms, and secondarily to vertical acceleration. Reducing input data to only those three vectors showed a very small decrease in the prediction performance. After the fine-tuning of the algorithm, a MAPE of 0.75% was obtained. Our results show that such very high performances can be expected even in multi-activity conditions and with low computational resource needs making this approach suitable for embedded devices.

Physical unclonable functions and QKD-based authentication scheme for IoT devices using blockchain

As the number of Internet of Things (IoT) devices is increasing exponentially, strong security measures are needed to guard against different types of cyberattacks. This research offers a novel IoT device authentication technique to mitigate these challenges by integrating three cutting-edge technologies namely blockchain technology, Quantum Key Distribution (QKD), and Physically Unclonable Functions (PUFs). By utilizing the distinctive qualities of PUFs for device identification and the unrivaled security of QKD for key exchange, the proposed approach seeks to address the significant security issues present in IoT environments. Adopting blockchain technology ensures transparency and verifiability of the authentication process across distributed IoT networks by adding an unchangeable, decentralized layer of trust. An examination of the computing and communication costs reveals that the proposed protocol is effective, necessitating low computational resources that are critical for IoT devices with limited resources. The protocol’s resistance against a variety of attacks is demonstrated by formal proofs based on the Real-Or-Random (ROR) model and security evaluations using the Scyther tool, ensuring the integrity and secrecy of communications. Various threats are analyzed, and the protocol is proven to be secure and efficient from all forms of attacks.

The Climate Data for Adaptation and Vulnerability Assessments and the Spatial Interactions Downscaling Method

This study presents the spatial interactions downscaling (SPID) method and introduces the climate data for adaptation and vulnerability assessments (ClimAVA) dataset. SPID employs random forest models to capture the relationship between spatial patterns at global circulation model (GCM) resolution and fine-resolution pixel values. In summary, a random forest model is trained for each fine spatial resolution pixel of the reference data as the predictand, and nine pixels from the spatially resampled (coarser) version of the reference data at the GCM’s resolutions as predictors. Models are then utilized to downscale the bias-corrected GCM data. The ClimAVA-SW dataset offers a high-resolution (4 km), bias-corrected, downscaled future climate projection derived from seventeen CMIP6 GCMs. It includes three variables (daily precipitation, minimum and maximum temperature) for three shared socioeconomic pathways (SSP245, SSP370, SSP585) across the U.S. Southwest region. The ClimAVA dataset sets itself apart with the SPID method’s capacity to provide remarkable climate realism, high physical plausibility of change, and excellent representation of extreme events while maintaining user-friendliness and requiring relatively low computational resources.

Methodology for 3D sound synthesis of directional acoustic sources by higher-order ambisonics

This paper presents the 3D sound field synthesis using both the multimodal method and higher-order ambisonics (HOA) of the pressure field radiated by directional acoustic sources. Ambisonics is a technique for encoding and reproducing measured or modeled (virtual) sound pressure field, based on a decomposition of the acoustic field over spherical harmonics. The directional source considered in this work is an acoustic horn excited by a flat piston. The free-field radiation from this horn is first modeled accurately over a wide frequency range using the multimodal method, which requires relatively low computational resources. This radiated pressure field, collected on a dual-layer sphere of virtual sensors distributed over a Lebedev geometry, allows its projection into the ambisonic domain. The pressure field is then synthetized in the laboratory's 3D 5th order HOA spatialization sphere, which consists of fifty-six loudspeakers. This offers the ability of listening to the radiated sound using a higher-order ambisonic synthesis of a 'virtual' source before it is manufactured. To qualitatively evaluate the performance of the proposed procedure, the transfer function of the synthesized horn is measured around the listening point within the spatialization sphere. Key words : 3D sound synthesis, high-order ambisonics, directive sources, waveguides, horn, multimodal method

Implicit unified gas-kinetic scheme for steady state solution of hypersonic thermodynamic non-equilibrium flows

The unified gas-kinetic scheme (UGKS), designed to solve Boltzmann equation and its model equations, aims to accurately resolve multi-scale flow problems within a single computation framework. In this study, an implicit UGKS is proposed for the steady state solution of diatomic gas flows based on a multiple-temperature Boltzmann model equation with non-equilibrium among translational, rotational and vibrational modes (Zhang et al., 2023), utilizing a novel hybrid Cartesian-unstructured discrete velocity space (DVS) mesh approach. The multiple-temperature macroscopic equations corresponding to the Boltzmann model equation are simultaneously solved implicitly to address the stiffness and nonlinearity of the collision operator encountered when solving the Boltzmann model equation in isolation. Both macroscopic equations and Boltzmann model equation are solved by using the point relaxation symmetric Gauss–Seidel method to achieve a fast convergence rate. The efficiency of the implicit UGKS can be improved by about two orders of magnitude compared to the explicit one. Moreover, the hybrid Cartesian-unstructured DVS achieves a noteworthy reduction in both CPU-hours and memory consumption, requiring only 15%∼23% of the advanced unstructured DVS. It effectively alleviates the dimensional crisis of UGKS in solving three-dimensional hypersonic flows, and can be extended to other DVS-based methods. As a result, UGKS extends its applicable regime to simulations of hypersonic thermodynamic non-equilibrium flows with Mach numbers up to 30, achieving three-dimensional flow simulations of complex geometrical configurations with relatively low computational resources. A series of test cases, including normal shock structures, two-dimensional hypersonic flows around cylinder, blunt wedge and flat plate, and three-dimensional hypersonic flows around a sphere and an X38-like configuration vehicle, are conducted to validate the implicit UGKS with hybrid Cartesian-unstructured DVS. Overall, the proposed method shows advantages of unified multi-scale methods in terms of computational efficiency and capturing multi-scale solution of hypersonic thermodynamic non-equilibrium flows.

Swimtrans Net: a multimodal robotic system for swimming action recognition driven via Swin-Transformer.

Currently, using machine learning methods for precise analysis and improvement of swimming techniques holds significant research value and application prospects. The existing machine learning methods have improved the accuracy of action recognition to some extent. However, they still face several challenges such as insufficient data feature extraction, limited model generalization ability, and poor real-time performance. To address these issues, this paper proposes an innovative approach called Swimtrans Net: A multimodal robotic system for swimming action recognition driven via Swin-Transformer. By leveraging the powerful visual data feature extraction capabilities of Swin-Transformer, Swimtrans Net effectively extracts swimming image information. Additionally, to meet the requirements of multimodal tasks, we integrate the CLIP model into the system. Swin-Transformer serves as the image encoder for CLIP, and through fine-tuning the CLIP model, it becomes capable of understanding and interpreting swimming action data, learning relevant features and patterns associated with swimming. Finally, we introduce transfer learning for pre-training to reduce training time and lower computational resources, thereby providing real-time feedback to swimmers. Experimental results show that Swimtrans Net has achieved a 2.94% improvement over the current state-of-the-art methods in swimming motion analysis and prediction, making significant progress. This study introduces an innovative machine learning method that can help coaches and swimmers better understand and improve swimming techniques, ultimately improving swimming performance.

Efficient Federated Learning Using Dynamic Update and Adaptive Pruning with Momentum on Shared Server Data

Despite achieving remarkable performance, Federated Learning (FL) encounters two important problems, i.e., low training efficiency and limited computational resources. In this paper, we propose a new FL framework, i.e., FedDUMAP, with three original contributions, to leverage the shared insensitive data on the server in addition to the distributed data in edge devices so as to efficiently train a global model. First, we propose a simple dynamic server update algorithm, which takes advantage of the shared insensitive data on the server while dynamically adjusting the update steps on the server in order to speed up the convergence and improve the accuracy. Second, we propose an adaptive optimization method with the dynamic server update algorithm to exploit the global momentum on the server and each local device for superior accuracy. Third, we develop a layer-adaptive model pruning method to carry out specific pruning operations, which is adapted to the diverse features of each layer so as to attain an excellent trade-off between effectiveness and efficiency. Our proposed FL model, FedDUMAP, combines the three original techniques and has a significantly better performance compared with baseline approaches in terms of efficiency (up to 16.9 times faster), accuracy (up to 20.4% higher), and computational cost (up to 62.6% smaller).

Comparative Analysis of CNN Algorithms for Mushroom Classification with Proposed Lightweight CNN Model

The classification of mushroom species presents significant ecologic and health-related challenges; advancement in classification techniques is required to gain reliable identifications. This study aims to explain a methodology that was devised and evaluated in the development of a novel, lightweight Convolutional Neural Network (CNN) designed specifically for the task of mushroom classification. The paper provides a custom CNN model that is computationally cost-effective and capable of high-precision classification, fit for real-time usage. Hence, the proposed model was evaluated on this dataset of curated mushroom images with traditional classifiers and state-of-the-art CNN architectures, such as EfficientNet-B7, ResNet50, InceptionV3, and MobileNetV2. The custom model is depth-wise separations engineered in such a way that while they reduce the computational load, they don't compromise the effectiveness of the model. The custom model achieved a test score of 0.68, which is moderate compared to more established models such as EfficientNet-B7 or ResNet50. This approach helps the model function effectively even on platforms having low computational resources. A comprehensive evaluation reveals that a custom CNN has reasonable accuracy in the identification of different mushroom species vis-à-vis existing models, but also significantly lightens the classification process.

Interpolation-split: a data-centric deep learning approach with big interpolated data to boost airway segmentation performance

The morphology and distribution of airway tree abnormalities enable diagnosis and disease characterisation across a variety of chronic respiratory conditions. In this regard, airway segmentation plays a critical role in the production of the outline of the entire airway tree to enable estimation of disease extent and severity. Furthermore, the segmentation of a complete airway tree is challenging as the intensity, scale/size and shape of airway segments and their walls change across generations. The existing classical techniques either provide an undersegmented or oversegmented airway tree, and manual intervention is required for optimal airway tree segmentation. The recent development of deep learning methods provides a fully automatic way of segmenting airway trees; however, these methods usually require high GPU memory usage and are difficult to implement in low computational resource environments. Therefore, in this study, we propose a data-centric deep learning technique with big interpolated data, Interpolation-Split, to boost the segmentation performance of the airway tree. The proposed technique utilises interpolation and image split to improve data usefulness and quality. Then, an ensemble learning strategy is implemented to aggregate the segmented airway segments at different scales. In terms of average segmentation performance (dice similarity coefficient, DSC), our method (A) achieves 90.55%, 89.52%, and 85.80%; (B) outperforms the baseline models by 2.89%, 3.86%, and 3.87% on average; and (C) produces maximum segmentation performance gain by 14.11%, 9.28%, and 12.70% for individual cases when (1) nnU-Net with instant normalisation and leaky ReLU; (2) nnU-Net with batch normalisation and ReLU; and (3) modified dilated U-Net are used respectively. Our proposed method outperformed the state-of-the-art airway segmentation approaches. Furthermore, our proposed technique has low RAM and GPU memory usage, and it is GPU memory-efficient and highly flexible, enabling it to be deployed on any 2D deep learning model.

Light-YOLO: A Study of a Lightweight YOLOv8n-Based Method for Underwater Fishing Net Detection

Detecting small dark targets underwater, such as fishing nets, is critical to the operation of underwater robots. Existing techniques often require more computational resources and operate under harsh underwater imaging conditions when handling such tasks. This study aims to develop a model with low computational resource consumption and high efficiency to improve the detection accuracy of fishing nets for safe and efficient underwater operations. The Light-YOLO model proposed in this paper introduces an attention mechanism based on sparse connectivity and deformable convolution optimized for complex underwater lighting and visual conditions. This novel attention mechanism enhances the detection performance by focusing on the key visual features of fishing nets, while the introduced CoTAttention and SEAM modules further improve the model’s recognition accuracy of fishing nets through deeper feature interactions. The results demonstrate that the proposed Light-YOLO model achieves a precision of 89.3%, a recall of 80.7%, and an mAP@0.5 of 86.7%. Compared to other models, our model has the highest precision for its computational size and is the lightest while maintaining similar accuracy, providing an effective solution for fishing net detection and identification.

Ultrafast jet classification at the HL-LHC

Three machine learning models are used to perform jet origin classification. These models are optimized for deployment on a field-programmable gate array device. In this context, we demonstrate how latency and resource consumption scale with the input size and choice of algorithm. Moreover, the models proposed here are designed to work on the type of data and under the foreseen conditions at the CERN large hadron collider during its high-luminosity phase. Through quantization-aware training and efficient synthetization for a specific field programmable gate array, we show that O(100) ns inference of complex architectures such as Deep Sets and Interaction Networks is feasible at a relatively low computational resource cost.

A thermal fluid dynamic model for the melt region during the laser powder bed fusion of polyamide 11 (PA11)

Studying the physical characteristics of the melt region in polymer-based laser powder bed fusion (PBF-LB/P) offers good comprehension that is essential for quality control, process optimization, understanding of material behavior, and preserving process consistency. Among the few numerical modelling investigations of PBF-LB/P, this study focuses on 3D multiphysics simulation of melt pool morphologies, beginning with powder bed preparation incorporating the actual particle size distribution using the discrete element method (DEM). A thermal fluid dynamic numerical model is then developed based on the finite volume method (FVM) using Flow-3D Weld to simulate PBF-LB while using polyamide 11 (PA11) as the primary material. The cooling after sintering plays a crucial role for the formation of the melt region due to the high viscosity of the polymer material effectively resulting in creeping flow. Here, we propose a model for both sintering as well as the subsequent cooling step, a two-step model capable of simulating the finished structure accurately with relatively low computational resources. The model's accuracy is verified through a mesh dependency analysis and its predictive power via validation against internal single-track experiments providing dimension data on the melt region width and depth. The simulation results allow for a detailed exploration of the effects of laser settings on melt region morphologies and the occurrence of defects. The model's scope is further extended to multi-track simulation, revealing the connection between surface roughness and laser settings. In general, this study provides significant contributions to the comprehension of PBF-LB/P methodologies which in turn can be used for further process improvement.

Prescribed Performance Spacecraft Attitude Control with Multiple Convergence Rates

This paper proposes a prescribed performance control policy for spacecraft attitude tracking control. The designed controller employs a kind of performance envelope function that can configure the convergence rate arbitrarily. Moreover, the controller design is based on the barrier Lyapunov function and indirect adaptive methods, which can suppress the fluctuation of external disturbances to obtain lower steady-state errors. Compared to prior studies on spacecraft attitude tracking, the proposed policy can arbitrarily configure the system convergence performance to achieve performance trade-offs in different scenarios. Moreover, the barrier Lyapunov function design and adaptive methods can achieve low steady-state errors with low computational resource consumption. Simulations that verify its effectiveness and superiority are provided herein.

Novel Deep Learning Domain Adaptation Approach for Object Detection Using Semi-Self Building Dataset and Modified YOLOv4

Moving object detection is a vital research area that plays an essential role in intelligent transportation systems (ITSs) and various applications in computer vision. Recently, researchers have utilized convolutional neural networks (CNNs) to develop new techniques in object detection and recognition. However, with the increasing number of machine learning strategies used for object detection, there has been a growing need for large datasets with accurate ground truth used for the training, usually demanding their manual labeling. Moreover, most of these deep strategies are supervised and only applicable for specific scenes with large computational resources needed. Alternatively, other object detection techniques such as classical background subtraction need low computational resources and can be used with general scenes. In this paper, we propose a new a reliable semi-automatic method that combines a modified version of the detection-based CNN You Only Look Once V4 (YOLOv4) technique and background subtraction technique to perform an unsupervised object detection for surveillance videos. In this proposed strategy, background subtraction-based low-rank decomposition is applied firstly to extract the moving objects. Then, a clustering method is adopted to refine the background subtraction (BS) result. Finally, the refined results are used to fine-tune the modified YOLO v4 before using it in the detection and classification of objects. The main contribution of this work is a new detection framework that overcomes manual labeling and creates an automatic labeler that can replace manual labeling using motion information to supply labeled training data (background and foreground) directly from the detection video. Extensive experiments using real-world object monitoring benchmarks indicate that the suggested framework obtains a considerable increase in mAP compared to state-of-the-art results on both the CDnet 2014 and UA-DETRAC datasets.

Iterative algorithm computational spectrometer based on a single-hidden-layer neural network.

Computational spectrometers have great application prospects in hyperspectral detection, and fast and high-precision in situ measurement is an important development trend. The computational spectrometer based on iterative algorithms has low requirements for computational resources and is easy to achieve hardware integration and in situ measurement. However, iterative algorithms are difficult to achieve high reconstruction accuracy due to the ill-posed nature of problems. Neural networks have powerful learning capabilities and can achieve high-precision spectral reconstruction. However, solely relying on neural network algorithms for reconstruction requires higher storage space and computing power from hardware devices, which makes it difficult to integrate large-scale neural network models into embedded systems. We propose using neural networks to alleviate the effect of the problem ill-posedness on the reconstruction results of iterative algorithms, so as to improve the reconstruction accuracy of the iterative algorithm computational spectrometers. First, spectral reconstruction was performed with iterative algorithms using a public spectral dataset. Then, a single-hidden-layer neural network was trained to establish a fitting relationship between the iterative algorithm spectral reconstruction results and the original spectrum. Finally, simulation and experimental results show that the proposed application of neural networks to alleviate the ill-posed problem of the iterative algorithm spectral reconstruction can effectively improve the reconstruction accuracy of iterative algorithm computational spectrometers with low computational resources. The research results may have good potential in achieving fast and high-precision in situ measurements of computational spectrometers.

EnParaNet: a novel deep learning architecture for faster prediction using low-computational resource devices

EnParaNet: a novel deep learning architecture for faster prediction using low-computational resource devices

Efficient blind super-resolution imaging via adaptive degradation-aware estimation

Single image super-resolution (SISR) has achieved prominent success based on deep learning. However, most SISR methods based on the specific degradation pattern, e.g., bicubic interpolation with a bulky network structure, are unsuitable for real-world images reconstructed on edge devices. To mitigate the above limitations, unlike the majority of blind super-resolution methods, which consist of a separate complex degradation kernel projection network and a SR reconstruction network, we propose an end-to-end lightweight multi-degradation oriented network that not only can estimate degradation kernel but also reconstructs HR image fast and accurately. The core components of the designed network adopt a lightweight SR model with region non-local feature similarity learning, i.e., Region Non-Local Feature Block (RNLFB), to establish region-wise global feature correlation. Concretely, more and fewer RNLFBs are used to learn more sophisticated feature representation and relatively simple degradation representation. Then, a degradation kernel projector is equipped to fulfill adaptive degradation-aware estimation from the degradation representation. With the help of the encoded degradation kernel, the reconstruction model can learn to restore a high-resolution image from the feature representation. By doing so, the degradation kernel estimation task and the HR image reconstruction task can be accomplished without designing two complex networks separately, which guides more stable and effective network training under low computational resource costs. Extensive experiments on synthetic and real-world datasets demonstrate that our method can fulfill precise degradation kernel prediction and HR image reconstruction compared with other state-of-the-art SR methods with lower model complexity.