Spiking Neural Networks (SNNs) offer a biologically inspired computational paradigm that emulates neuronal activity through discrete spike-based processing. Despite their advantages, training SNNs with traditional backpropagation (BP) remains challenging due to computational inefficiencies and a lack of biological plausibility. This study explores the Forward-Forward (FF) algorithm as an alternative learning framework for SNNs. Unlike backpropagation, which relies on forward and backward passes, the FF algorithm employs two forward passes, enabling layer-wise localized learning, enhanced computational efficiency, and improved compatibility with neuromorphic hardware. We introduce an FF-based SNN training framework and evaluate its performance across both non-spiking (MNIST, Fashion-MNIST, Kuzushiji-MNIST) and spiking (Neuro-MNIST, SHD) datasets. Experimental results demonstrate that our model surpasses existing FF-based SNNs on evaluated static datasets with a much lighter architecture while achieving accuracy comparable to state-of-the-art backpropagation-trained SNNs. On more complex spiking tasks such as SHD, our approach outperforms other SNN models and remains competitive with leading backpropagation-trained SNNs. These findings highlight the FF algorithm's potential to advance SNN training methodologies by addressing some key limitations of backpropagation.

Mixed Integer Linear Programming (MILP) is utilized in behavioral synthesis as a mathematical model to design efficient hardware. However, solving large MILP models poses significant computational challenges due to their NP-hard nature. Paralleling can tackle this challenge by amortizing the execution time, yet unbalanced loads can hinder its effectiveness. In this paper, we address the load balance issue of parallel Branch and Bound (B&B) algorithms, particularly sub-tree parallelism, which exhibit efficiency in solving MILP models derived from behavioral synthesis. The proposed algorithm strategically partitions the original problem into sub-problems by selecting decision variables that appear in a higher number of constraints to prioritize load balance and enhance solver performance. We evaluate the effectiveness of our method using MILP models derived from Mediabench data flow graphs of various sizes. The experimental results indicate that the proposed algorithm achieves speedups ranging from approximately 1 to 13 times, highlighting its efficacy in improving the scalability and efficiency of MILP solving for behavioral synthesis.

The structural change toward the digital transformation of online sales elevates the importance of parallel processing techniques in recommender systems, particularly in the pandemic and post-pandemic era. Matrix factorization (MF) is a popular and scalable approach in collaborative filtering (CF) to predict user preferences in recommender systems. Researchers apply Stochastic Gradient Descent (SGD) as one of the most famous optimization techniques for MF. Paralleling SGD methods help address big data challenges due to the wide range of products and the sparsity in user ratings. However, these methods' convergence rate and accuracy are affected by the dependency between the user and item latent factors, specifically in large-scale problems. Besides, the performance is sensitive to the applied learning rates. This article proposes a new parallel method to remove dependencies and boost speed-up by using fractional calculus to improve accuracy and convergence rate. We also apply adaptive learning rates to enhance the performance of our proposed method. The proposed method is based on Compute Unified Device Architecture (CUDA) platform. We evaluate the performance of our proposed method using real-world data and compare the results with the close baselines. The results show that our method can obtain high accuracy and convergence rate in addition to high parallelism.

Artificial Intelligence (AI) has emerged as a pivotal technology across various sectors, including healthcare, trans-portation, and the development of smart cities, revolutionizing service delivery and operational efficiency. However, the adoption and introduction of new data-driven services leveraging central-ized training models have been hindered by significant concerns over privacy and data security, as these traditional techniques potentially expose sensitive information to breaches. Federated Learning (FL) presents a compelling solution to this dilemma, enabling decentralized data processing without compromising privacy. Integrating Edge AI into this framework, FL enables the collaborative training of models based on data distributed across different clients. Nevertheless, implementing FL on edge devices introduces a set of challenges due to the limited computational and memory resources available on such tiny devices. Specifically, the backpropagation (BP) phase of training models is notably resource-intensive, posing a barrier to efficient deployment. To address this, we replaced the backpropagation phase with a Forward-Forward (FF) algorithm. Moreover, we integrated and compared several loss functions, namely Hinton, Symba, and Swish, to assess their compatibility and efficiency in the context of forward-forward training within the federated learning framework. The study indicates that our novel method leads to a slight decrease in accuracy for large and complex datasets compared to the traditional BP technique. However, it has the potential to enhance runtime and reduce memory overhead. The proposed technique represents a promising path toward the broader adoption of Edge AI by effectively addressing critical technical challenges, namely privacy concerns and on-chip model training.

Flying ad hoc networks (FANETs) have particular importance in various military and civilian applications due to their specific features, including frequent topological changes, the movement of drones in a three-dimensional space, and their restricted energy. These features have created challenges for designing cluster-based routing protocols. In this paper, a Q-learning-based smart clustering routing method (QSCR) is suggested in FANETs. In QSCR, each node discovers its neighbors through the periodic exchange of hello messages. The hello time interval is different in each cluster, and cluster leaders determine this interval based on the average speed similarity. Next, an adaptive clustering process is presented for categorizing drones in the clusters. In this step, the cluster leader is selected based on a new parameter called merit value, which includes residual energy, centrality, neighbor degree, speed similarity, and link validity time. Then, a centralized Q-learning model is presented to tune weight coefficients related to merit parameters dynamically. In the last step, the routing process is done using a greedy forwarding technique. Finally, QSCR is run on NS2, and the simulation results of QSCR are compared with those of ICRA, WCA, and DCA. These results show that QSCR carries out the clustering process rapidly but has less cluster stability than ICRA. QSCR gets energy efficiency and improves network lifetime. In the routing process, QSCR has a high packet delivery rate compared to other schemes. Also, the number of isolated clusters created in QSCR is less than other clustering methods. However, the proposed scheme has a higher end-to-end delay than ICRA. Also, this scheme experiences more communication overhead than ICRA slightly.

Network Intrusion Detection Systems (NIDSes) are essential for safeguarding critical information systems. However, the lack of adaptability of Machine Learning (ML) based NIDSes to different environments could cause slow adoption. In this paper, we propose a multimodal NIDS that combines flow and payload features to detect cyber-attacks. The focus of the paper is to evaluate the use of multimodal traffic features in detecting attacks, but not on a practical online implementation. In the multimodal NIDS, two random forest models are trained to classify network traffic using selected flow-based features and the first few bytes of protocol payload, respectively. Predictions from the two models are combined using a soft voting approach to get the final traffic classification results. We evaluate the multimodal NIDS using flow-based features and the corresponding payloads extracted from Packet Capture (PCAP) files of a publicly available UNSW-NB15 dataset. The experimental results show that the proposed multimodal NIDS can detect most attacks with average Accuracy, Recall, Precision and F1 scores ranging from 98% to 99% using only six flow-based traffic features, and the first 32 bytes of protocol payload. The proposed multimodal NIDS provides a reliable approach to detecting cyber-attacks in different environments.

An important issue in data mining and information retrieval is the problem of multiple similarity or n-similarity. This problem entails finding a group of n data points with the highest similarity within a large dataset. Exact methods to solve this problem exist but come with high time and space complexities. Additionally, various metaheuristic algorithms have been proposed, including genetic algorithms, gravitational search algorithms, particle swarm optimization, imperialist competitive algorithms, and fuzzy imperialist competitive algorithms. These metaheuristics are capable of finding near-optimal solutions within a reasonable timeframe, although there is no guarantee of achieving exact results. In this paper, we employ a parallelization technique using CUDA to expedite the exact method. We conduct experiments on textual datasets to identify a group of n textual documents with the highest similarity to each other. The experimental results demonstrate that the proposed parallel exact method significantly reduces execution time compared to the best sequential approach and CPU multi-core implementation. Furthermore, it is evident that the proposed method requires less memory space than the exact method.

Recognizing musical instruments from an audio signal is a challenging yet valuable endeavor within the realm of music study. The recognition and classification of musical instruments could prove beneficial in organizing various genres of music and automation in music transcription reading and producing. In this paper, we will investigate the use of a Deep Convolutional Neural Network for automatic instruments recognition of polyphonic music. We enhance the state-of-the-art model, which we establish to be responsive to the instrument's playing style rather than its timbre. Furthermore, we set up experimental validation on small networks to extract timbre features from a spectrogram. We demonstrate an ensemble model based on these experiments, which improves the model accuracy by 20 % for both single and multiple instrument recognition. Additionally, we present several models capable of achieving competitive performance with a significantly smaller number of network parameters and neurons.

Motor imagery (MI) is a widely used technique in brain-computer interface (BCI) systems, which allows users to control external devices using their brain signals. Electroen-cephalogram (EEG) signals are commonly used to detect and classify MI tasks. However, the lack of annotated data hampers the performance of machine learning (ML) algorithms. Federated learning (FL) is a promising approach to address these challenges, as it allows models to be trained on decentralized datasets without exchanging data. This study proposes an FL approach for MI-EEG signal classification using a convolutional neural network (CNN) on the PhysioNet dataset containing EEG recordings of left and right-hand imagery movements. We evaluate the performance of the FL approach using two different aggregation methods, namely FedAvg and FedProx, and compare it to the centralized ML approach. Furthermore, we explore the effect of increasing the number of clients who participate in the learning process on the performance of the model. Our findings demonstrate that FL maintains consistent classification accuracy comparable to the centralized ML approach while reducing data leakage risks. Thus, FL shows great promise as an essential instrument for MI-EEG signal classification and BCI systems, enabling distributed training of a comprehensive model when privacy-sensitive data is dispersed across multiple clients.

In this paper, we present two distinct hardware dynamic memory allocation schemes that are based on the binary buddy system algorithm. Our aim is to mitigate internal fragmentation without impacting the area and performance of the system. The first scheme introduces a parallel design for calculating the addresses of free blocks, which results in a decrease in allocation latency while maintaining acceptable resource utilization. This scheme is particularly well-suited for managing a limited number of minimum allocable units (MAU). On the other hand, the second allocator can handle a large number of MAUs due to its innovative searching mechanism. This allocator exhibits lower resource consumption and operates with an acceptable allocation latency. Furthermore, to decrease internal fragmentation, we develop a novel update mechanism for allocating information data structures in both methods. By employing these two allocator schemes, we can improve the efficiency and resource management of dynamic memory allocation for hardware systems. Experimental results demonstrate that the first and second proposed schemes achieve a minimum allocation speed-up of $\times 2$ and $\times 1.8$ compared to their counterparts. At the same time, they achieve a reduction of at least 78% and %88 in resource utilization, respectively. The results show that the total fragmentation is reduced by at least 14% due to the lower internal fragmentation.

Modeling high-level synthesis (HLS) as mixed integer linear programming (MILP) affords the opportunity to integrate constraints and optimization objectives of hardware design in the form of a mathematical intermediate representation. Consequently, it is possible to improve previously developed methods to solve MILP models and customize them for application in domain-specific functions. However, the problem remains NP-hard, and solving large models requires methods that investigate the state space intelligently. Despite the high potential of branch and bound (B&B) algorithms to solve MILP models quickly, computing the best answer is still an open challenge. In this paper, we first develop three model improvement techniques that reduce the size of the search space for MILP models derived from HLS. Then, we present a new B&B algorithm to tackle the computational challenge by considering the properties of the original HLS problem. In this regard, we propose two heuristic techniques that enable the B&B algorithm to prove that some branches are not promising and should be pruned. Moreover, we have developed a best-first strategy by suggesting a new priority key calculation scheme to improve tree traversal in the B&B algorithm. Besides, the sifting method has been employed as the LP relaxation method to achieve fast processing of large models. Our proposed approach was evaluated using a set of MILP models derived from the synthesis of Mediabench data flow graphs. The experimental results indicate that our approach outperforms modern MILP solvers in terms of speed and the scale of the MILP models which can solve. According to the test results on the large MILP models, we solved models with 7254 integer and binary variables in less than 13 minutes.

Image encryption is a reliable method for securely transmitting images over a network. The time required to encrypt and decrypt an image in online applications is also very important. Although cellular automata cryptography is an appropriate technique for parallelizing and accelerating cryptographic methods, its capacity cannot be demonstrated only in multi-core platforms. Thus, it is needed to parallelize cellular automata cryptography on Graphic Processor Units (GPUs) in order to significantly decrease the encryption/decryption time. In this paper, we propose a new parallel algorithm for two-dimensional cellular automata cryptography that is implemented on GPU. The proposed algorithm uses multiple threads at once to accelerate the bit-level permutation and substitution operations by taking into account the capacity of cellular automata in parallel processing. According to the study experimental findings, the proposed algorithm performs faster on GPU compared to a multicore platform while maintaining the same level of security in comparison to the serial algorithm.

Recommender systems enable companies to generate meaningful recommendations to users for items or products that might interest them. Stochastic Gradient Descent Matrix Factorization (SGD-MF) is one of the most popular model-based recommender systems. Fractional Adaptive Stochastic Gradient Descent matrix factorization (FASGD-MF) is a subset of SGD-MF-based models that apply fractional calculus in an adaptive way. There are some hyperparameters in these models that impact the quality of the recommender system. However, searching the hyperparameter space to find the best configuration using an exhaustive search is often a time-consuming task. This paper employs a genetic algorithm as a search metaheuristic to tackle this problem. The proposed method is designed based on non-uniform mutation and whole arithmetic crossover. The results indicate that optimizing hyperparameters by the proposed method not only adjusts the values of hyperparameters automatically but also can improve the quality of SGD-MF-based models. Implementing the proposed genetic algorithm on two datasets (MovieLens 100K and MovieLens 1M) verifies the assertion about the performance.

Fabricating of metal foams with desired morphological parameters including pore size, porosity and pore opening is possible now using sintering technology. Thus, if it is possible to determine the morphology of metal foam to absorb sound at a given frequency, and then fabricate it through sintering, it is expected to have optimized metal foams for the best sound absorption. Theoretical sound absorption models such as Lu model describe the relationship between morphological parameters and the sound absorption coefficient. In this study, the Lu model was used to optimize the morphological parameters of Aluminum metal foam for the best sound absorption coefficient. For this purpose, the Lu model was numerically solved using written codes in MATLAB software. After validating the proposed codes with benchmark data, the genetic algorithm (GA) was applied to optimize the affecting morphological parameters on the sound absorption coefficient. The optimization was carried out for the thicknesses of 5 mm to 40 mm at the sound frequency range of 250 Hz–8000 Hz. The optimized parameters ranged from 50% to 95% for porosity, 0.1 mm to 4.5 mm for pore size, and 0.07 mm to 0.6 mm for pore opening size. The result of this study was applied to fabricate the desired Aluminum metal foams for the best sound absorption. The novel approach applied in this study, is expected to be successfully applied in for best sound absorption in desired frequencies.

Solving Integer Linear Programming (ILP) models generally lies in the category of NP-hard problems and finding the optimal answer for large models is a computational challenge. Genetic algorithms are a family of metaheuristic algorithms capable of adjusting and redesigning parameters and operations according to the characteristics of ILP models. On the other hand, still the genetic algorithm performs a lot of operations to solve large models, and parallel processing is a suitable technique to tackle this problem. This paper introduces an LP-Relaxation based parallel genetic algorithm that uses a population-based incremental learning technique to presents an expandable solver for large ILP models derived from a behavioral synthesis of digital circuits. In the proposed algorithm, each chromosome provides a state subspace of possible solutions, and each generation is produced based on a probability vector as well as elitism. Our experiments verify the efficiency of the proposed algorithm on multicore platforms, as it outperformed four previous genetic algorithms for solving mixed integer programming problems. The proposed genetic algorithm solved 20 ILP models include up to 5183 int / binary decision variables in less than 20 min using four 16-core AMD Opteron 6386 SE processors. Also, the results indicate that for models with more than 4000 variables, the speedup and the efficiency of the proposed parallel genetic algorithm on 60 CPU cores is more than 18X and 30%, respectively.

Sensor nodes in a wireless sensor Network are assigned for different operational modes to perfume application-specific objectives. The decision to assign operational modes to nodes is a challenging problem in the presence of multiple criteria including energy-efficient, maintaining network connectivity, and fulfilling application goals. Several metaheuristic methods are introduced in the literature to address this NP-hard problem, however, these methods require further improvements in execution-time and finding the optimum solution. In this research, we propose an improved version of a metaheuristic method called Quadri-valent Quantum-Inspired Gravitational Search Algorithm (QQIGSA) to solve Quadri-valent problems by applying a Not Q-Gate and paralleling QQIGSA method on the graphics processing unit. The proposed method employs a heterogeneous platform and justifies its parameters. The experimental results show that the performance enhancement from 1.8 to 2.25 compared to the previous parallel implementations. Moreover, we achieve the speedup of 8 by using the proposed heterogeneous paralleling technique.

Integer Linear Programming (ILP) formulation of behavioral synthesis allows hardware designers to implement efficient circuits considering resource and timing constraint. However, finding the optimal answer of ILP models is an NP-Hard problem and remains a computational challenge. In this paper, we address this challenge by developing two exact parallel branch and bound algorithms which are capable of solving large-scale ILP models derived from behavioral synthesis. The first algorithm enables sub-node parallelism as well as adaptive branching and memory efficient techniques to accelerate solving ILP models on shared memory multi-core systems. The second algorithm is developed based on node parallelism strategy. We evaluated the proposed algorithms using large ILP models derived from Media Bench Data Flow Graphs. The experimental results indicate both the proposed methods can successfully accelerate behavioral synthesis on multi-core platforms and outperforms IBM ILOG CPLEX (v12.60) MIP solver in solving large ILP models.

The Louvain community detection algorithm is a hierarchal clustering method categorized in the NP-hard problem. Its execution time to find communities in large graphs is, therefore, a challenge. Parallelization is an effective solution for amortizing Louvain's execution time. In this paper, we propose an adaptive CUDA Louvain method (ACLM) algorithm that benefits from the graphic processing unit (GPU). ACLM uses the shared memory in GPU, as well as the optimal number of threads in the GPU blocks. These features minimize parallelization overhead and accelerate the calculation of modularity parameters. The proposed algorithm allocates threads to each block based on the number of required streaming multiprocessors (SMs) and warps on GPU. The implementation results show that ACLM can effectively accelerate the execution time by 77% compared to the competitive method in the large graph benchmarks.

For reconstructing CT images in the clinical setting, ‘effective energy’ is usually used instead of the total X-ray spectrum. This approximation causes an accuracy decline. We proposed to quantize the total X-ray spectrum into irregular intervals to preserve accuracy. A phantom consisting of the skull, rib bone, and lung tissues was irradiated with CT configuration in GATE/GEANT4. We applied inverse Radon transform to the obtained Sinogram to construct a Pixel-based Attenuation Matrix (PAM). PAM was then used to weight the calculated Hounsfield unit scale (HU) of each interval's representative energy. Finally, we multiplied the associated normalized photon flux of each interval to the calculated HUs. The performance of the proposed method was evaluated in the course of Complexity and Visual analysis. Entropy measurements, Kolmogorov complexity, and morphological richness were calculated to evaluate the complexity. Quantitative visual criteria (i.e., PSNR, FSIM, SSIM, and MSE) were reported to show the effectiveness of the fuzzy C-means approach in the segmenting task.

Background and Objectives: -similarity problem defined as measuring the similarity among objects and finding a group of objects from a dataset that have the most similarity to each other. This problem has been become an important issue in information retrieval and data mining. Theory of this concept is mathematically proven, but it practically has high memory complexity and is so time consuming. Besides, the solutions found by metaheuristics are not exact.Methods: This paper is conducted to propose an exact method to solve -similarity problem reducing the memory complexity and decreasing the execution time by parallelism using Open-MP. The experiments are performed on the application of text document resemblance.Results: It has been shown that the memory complexity of the proposed method is decreased to , and the experimental results show that this method accelerates the speed of the computations about 5 times.Conclusion: The simulated results of the proposed method display a good improvement in speed, the used memory space, and scalability compared with the previous exact method.