We present a C++14 library for performance portability of scientific computing codes across CPU and GPU architectures. Our library combines generic data structures like vectors, multi-dimensional arrays, maps, graphs,...
详细信息
Suicide continues to be an issue in our society. Studies agree that it is best to deal with suicidal ideation in its early stages, and for this reason, researchers have been conducting experiments training different N...
详细信息
ISBN:
(数字)9798331530891
ISBN:
(纸本)9798331530907
Suicide continues to be an issue in our society. Studies agree that it is best to deal with suicidal ideation in its early stages, and for this reason, researchers have been conducting experiments training different Neural Networks (NN) to detect it. Transformers are the dominant Neural Network architecture in the domain of Suicidal Ideation detection, being a robust solution for not only the proposed problem but for a wide variety of NLP problems. LSTM-CNN, one of the prominent architectures in the field is also proposed as a great solution. this study aims to evaluate the performance of BERT, RoBERTa, LSTM-CNN, and Bi-LSTM-CNN models for suicidal ideation detection. Our experiments indicated that BERT models have an edge over both LSTM-CNN and BI-LSTM-CNN models, scoring up to 0.986 accuracy on our test set. Furthermore, while directly comparing LSTM-CNN with Bi-LSTM-CNN, it was observed that the difference between the models isn’t significant. Our paper contributes to the domain by proving no advantage of using LSTM-CNN models over the Transformers.
In high-speed arithmetic circuits, adder efficiency is crucial for system performance. Common adders like Carry Skip, Han-Carlson, and Kogge Stone aim to optimize delay and area but still struggle to balance speed, ar...
详细信息
ISBN:
(数字)9798350379884
ISBN:
(纸本)9798350379891
In high-speed arithmetic circuits, adder efficiency is crucial for system performance. Common adders like Carry Skip, Han-Carlson, and Kogge Stone aim to optimize delay and area but still struggle to balance speed, area, and power. this paper introduces a new three-operand pipelined adder, replacing conventional parallel prefix logic with a streamlined pipelined method. the design combines an adder and subtractor in one unit, using modular logic to handle both functions. Pipelining allows continuous data processing, increasing throughput and reducing critical path delay. this three-operand pipelined adder achieves significant area and delay improvements over traditional designs. Using fewer logic gates and connections makes the design more compact and improves its performance. Simulations and synthesis validate the proposed design's effectiveness, showing major improvements in speed and area, making it valuable for high-speed, area-limited VLSI applications.
A multi-channel parallel convolutional neural network model is proposed to predict the mechanical properties of hot-rolled steel by chemical content and process parameters. the innovative contribution of this paper is...
ISBN:
(纸本)9798400716485
A multi-channel parallel convolutional neural network model is proposed to predict the mechanical properties of hot-rolled steel by chemical content and process parameters. the innovative contribution of this paper is to propose an improved Gramian Angular field method in the field of steel performance prediction, which converts the original data into a two-dimensional image matrix and introduces new features. Compared withthe traditional convolutional neural network model, the multi-channel parallel convolutional neural network uses private blocks to decouple data and public blocks to integrate key data, which effectively improves the prediction accuracy of the model. Experiments in this paper show that the optimal hyperparameter structure model proposed in this paper has higher prediction accuracy than similar models proposed in other literatures.
Advancement in technology leads to connecting different types of devices or things to the Internet and enables the formation of a special kind of network called the Internet of things (IoT). Intrusion detection in an ...
详细信息
ISBN:
(纸本)9781665476485
Advancement in technology leads to connecting different types of devices or things to the Internet and enables the formation of a special kind of network called the Internet of things (IoT). Intrusion detection in an IoT is a challenging task due to its unique characteristics. Machine learning schemes possess the potential to improve intrusion detection systems in case of an IoT. In this paper, we present a survey of advancements in research on the use of machine learning approaches for intrusion detection in an IoT. Our focus is on architectures, schemes, and the types of machine learning approaches used for intrusion detection. We compare different schemes based on their basis and features.
Computer vision (CV) based inspection has recently attracted considerable attention and is progressively replacing traditional visual inspection which is subject to poor accuracy, high subjectivity and inefficiency. T...
详细信息
ISBN:
(纸本)9781450385893
Computer vision (CV) based inspection has recently attracted considerable attention and is progressively replacing traditional visual inspection which is subject to poor accuracy, high subjectivity and inefficiency. this paper, benefiting from hybrid structures of multichannel parallel convolutional neural networks (pCNNs), introduces a unique deep learning framework for road crack detection. Ideally, CNN-based frameworks require a relatively huge computing resources for accurate image analysis. However, the portability objective of this work necessitates the utilization of low power processing units. To that purpose, we propose robust deep representation learning for Road Crack Detection (RoCDe) which uses a multichannel pCNNs. Bayesian optimization algorithm (BOA) was used to optimize the multichannel pCNNs training withthe fewest possible neural network (NN) layers to achieve the maximum accuracy, improved efficiency and minimum processing time. the CV training was done using two distinct optimizers namely Adam and RELU on a sufficiently available datasets through image preprocessing and data augmentation. Experimental results show that, the proposed algorithm can achieve high accuracy around 95% in crack detection, which is good enough to replace human inspections normally conducted on-site. this is largely due to well-calibrated predictive uncertainty estimates (WPUE). Effectiveness of the proposed model is demonstrated and validated empirically via extensive experiments and rigorous evaluation on large scale real world datasets. Furthermore, the performance of hybrid CNNs is compared with state-of-the art NN models and the results provides remarkable difference in success level, proving the strength of multichannel pCNNs.
the reactive force field (ReaxFF) interatomic potential is a powerful tool for simulating the behavior of molecules in a wide range of chemical and physical systems at the atomic level. Unlike traditional classical fo...
详细信息
the proceedings contain 29 papers. the special focus in this conference is on Machine Learning, Image processing, Network Security and Data Sciences. the topics include: the Potential of 1D-CNN for EEG Menta...
ISBN:
(纸本)9783031622168
the proceedings contain 29 papers. the special focus in this conference is on Machine Learning, Image processing, Network Security and Data Sciences. the topics include: the Potential of 1D-CNN for EEG Mental Attention State Detection;marker-Based Augmented Reality Application in Education Domain;detection and Classification of Waste Materials Using Deep Learning Techniques;Deep Learning Based EV’s Charging Network Management;internet of Medical things: Empowering Mobility and Health Monitoring with a Smart Walking Stick;SynText - Data Augmentation Algorithm in NLP to Improve Performance of Emotion Classifiers;preface;Phishing Detection Using 1D-CNN and FF-CNN Models Based on URL of the Website;A Comparative Analysis of ML Based Approaches for Identifying AQI Level;a Review of Authentication Schemes in Internet of things;advancements in Facial Expression Recognition: A Comprehensive Analysis of Techniques;a Deep Learning Method for Obfuscated Android Malware Detection;COVID-19 Detection from Chest X-Ray Images Using GBM with Comparative Analysis;hate Speech Detection Using Machine Learning and Deep Learning Techniques;effectiveness of Influencer Marketing on Gen Z Consumers;compressors Using Modified Sorting and parallel Counting;violence Detection in Indoor Domestic Environment Using Multimodal Information;Code-Mixed Language Understanding Using BiLSTM-BERT Multi-attention Fusion Mechanism;diabetes Prediction Using Machine Learning Classifiers;crop Yield Prediction Using Machine Learning Approaches;Comparative Analysis of Economy-Based Multivariate Oil Price Prediction Using LSTM;MRI Based Spatio-Temporal Model for Alzheimer’s Disease Prediction;bridging the Gap: Condensing Knowledge Graphs for Metaphor processing by Visualizing Relationships in Figurative and Literal Expressions;a Novel Unsupervised Learning Approach for False Data Injection Attack Detection in Smart Grid;a Multi-stage Encryption Technique Using Asymmetric and Various Symmetric Ciphers;speed-Invar
Cloud-edge collaborative computing paradigm is a promising solution to high-resolution video analytics systems. the key lies in reducing redundant data and managing fluctuating inference workloads effectively. Previou...
详细信息
ISBN:
(数字)9798350386059
ISBN:
(纸本)9798350386066
Cloud-edge collaborative computing paradigm is a promising solution to high-resolution video analytics systems. the key lies in reducing redundant data and managing fluctuating inference workloads effectively. Previous work has focused on extracting regions of interest (RoIs) from videos and transmitting them to the cloud for processing. However, a naive Infrastructure as a Service (IaaS) resource configuration falls short in handling highly fluctuating workloads, leading to violations of Service Level Objectives (SLOs) and inefficient resource utilization. Besides, these methods neglect the potential benefits of RoIs batching to leverage parallelprocessing. In this work, we introduce Tangram, an efficient serverless cloud-edge video analytics system fully optimized for both communication and computation. Tangram adaptively aligns the RoIs into patches and transmits them to the scheduler in the cloud. the system employs a unique “stitching” method to batch the patches with various sizes from the edge cameras. Additionally, we develop an online SLO-aware batching algorithm that judiciously determines the optimal invoking time of the serverless function. Experiments on our prototype reveal that Tangram can reduce bandwidth consumption and computation cost up to 74.30 % and 66.35 %, respectively, while maintaining SLO violations within 5 % and the accuracy loss negligible.
In this paper we present a parallel version for the algorithm BFQzip, we introduced in [Guerrini et al., BIOSTEC – BIOINFORMATICS 2022], that modifies the bases and quality scores components taking into account both ...
详细信息
暂无评论