The rapidly advancing Convolutional Neural Networks(CNNs)have brought about a paradigm shift in various computer vision tasks,while also garnering increasing interest and application in sensor-based Human Activity Rec...
详细信息
The rapidly advancing Convolutional Neural Networks(CNNs)have brought about a paradigm shift in various computer vision tasks,while also garnering increasing interest and application in sensor-based Human Activity Recognition(HAR)***,the significant computational demands and memory requirements hinder the practical deployment of deep networks in resource-constrained *** paper introduces a novel network pruning method based on the energy spectral density of data in the frequency domain,which reduces the model’s depth and accelerates activity *** traditional pruning methods that focus on the spatial domain and the importance of filters,this method converts sensor data,such as HAR data,to the frequency domain for *** emphasizes the low-frequency components by calculating their energy spectral density ***,filters that meet the predefined thresholds are retained,and redundant filters are removed,leading to a significant reduction in model size without compromising performance or incurring additional computational ***,the proposed algorithm’s effectiveness is empirically validated on a standard five-layer CNNs backbone *** computational feasibility and data sensitivity of the proposed scheme are thoroughly ***,the classification accuracy on three benchmark HAR datasets UCI-HAR,WISDM,and PAMAP2 reaches 96.20%,98.40%,and 92.38%,***,our strategy achieves a reduction in Floating Point Operations(FLOPs)by 90.73%,93.70%,and 90.74%,respectively,along with a corresponding decrease in memory consumption by 90.53%,93.43%,and 90.05%.
Since most multiobjective optimization problems in real-world applications contain constraints, constraint-handling techniques (CHTs) are necessary for a multiobjective optimizer. However, existing CHTs give no relaxa...
详细信息
We consider the online convex optimization (OCO) problem with quadratic and linear switching cost when at time t only gradient information for functions fτ, τ 16(Lµ+5) for the quadratic switching cost, and also...
详细信息
We consider the online convex optimization (OCO) problem with quadratic and linear switching cost when at time t only gradient information for functions fτ, τ 16(Lµ+5) for the quadratic switching cost, and also show the bound to be order-wise tight in terms of L, µ. In addition, we show that the competitive ratio of any online algorithm is at least max{Ω(L), Ω(pLµ )} when the switching cost is quadratic. For the linear switching cost, the competitive ratio of the OMGD algorithm is shown to depend on both the path length and the squared path length of the problem instance, in addition to L, µ, and is shown to be order-wise, the best competitive ratio any online algorithm can achieve. Copyright is held by author/owner(s).
In order to dynamically create a sequence of textual descriptions for images, image description models often make use of the attention mechanism, which involves an automatic focus on different regions within an image....
详细信息
This article aims to improve the detection of ASD in children using ET data and advanced ML techniques. ASD, a neurodevelopmental disorder characterized by impairments in social communication, interaction, and repetit...
详细信息
This article aims to improve the detection of ASD in children using ET data and advanced ML techniques. ASD, a neurodevelopmental disorder characterized by impairments in social communication, interaction, and repetitive behaviors, typically manifests before the age of three. Early diagnosis is crucial for timely and effective interventions. ET data have revealed distinct gaze patterns in children with ASD, such as diminished attention to social cues and increased fixation on repetitive stimuli. However, current methods of ET data analysis are largely manual and lack scalability for widespread clinical screening. To address these limitations, we propose a novel, scanpath-based ASD detection method that identifies atypical gaze behaviors through dynamic changes in visual attention. We extract four sequential features from scanpaths and analyze variations in feature space across different ASD severity levels low, mild, moderate, and severe using the MultiMatch and DTW similarity metrics. Our analysis reveals that children with ASD exhibit unique and highly individualized gaze patterns when compared to TD children. Notable differences are observed in attention duration and vertical gaze distribution, providing key insights into ASD-related visual behaviors. For classification, we employ a hybrid CNN-RNN model, which significantly outperforms traditional ML methods. The CNN-RNN model achieves an accuracy of 97%, recall of 98.24%, and an F1-score of 97.04% using the feature set (x, d, y). In comparison, models based on GRU and 2-LSTM networks show competitive accuracies of 92% and 90%, respectively. However, RFC and XGBoost models underperform, with accuracies ranging between 70.25 and 80.80%. These findings demonstrate the efficacy of DL approaches, particularly the CNN-RNN hybrid model, in accurately classifying ASD based on ET data, emphasizing their potential to enhance diagnostic accuracy. The proposed scalable method for ASD detection holds promise for improving ea
Machine learning has become important for anomaly detection in water quality prediction. Data anomalies are often caused by the difficulties of analysing large amounts of data, both technical and human, but approaches...
详细信息
Surface Electromyographic (sEMG) signals are a promising approach to hand and finger gesture recognition. Most of the sEMG-based hand gesture recognition has developed based on the whole hand gesture, full wavelength,...
详细信息
Deep learning offers a promising methodology for the registration of prostate cancer images from histopathology to MRI. We explored how to effectively leverage key information from images to achieve improved end-to-en...
详细信息
1 Introduction In Natural Language Processing(NLP),topic modeling is a class of methods used to analyze and explore textual corpora,i.e.,to discover the underlying topic structures from text and assign text pieces to ...
详细信息
1 Introduction In Natural Language Processing(NLP),topic modeling is a class of methods used to analyze and explore textual corpora,i.e.,to discover the underlying topic structures from text and assign text pieces to different *** NLP,a topic means a set of relevant words appearing together in a particular pattern,representing some specific *** is beneficial for tracking social media trends,constructing knowledge graphs,and analyzing writing *** modeling has always been an area of extensive research in *** methods like Latent Semantic Analysis(LSA)and Latent Dirichlet Allocation(LDA),based on the“bag of words”(BoW)model,often fail to grasp the semantic nuances of the text,making them less effective in contexts involving polysemy or data noise,especially when the amount of data is small.
Deploying Unmanned Aerial Vehicles (UAVs) as aerial base stations enhances the coverage and performance of communication networks in Vehicular Edge Computing (VEC) scenarios. However, due to the limited communication ...
详细信息
暂无评论