检索结果-内蒙古大学图书馆

Advancing mental health detection in texts via multi-task learning with soft-parameter sharing transformers

Neural Computing and Applications 2025年第5期37卷 3077-3110页

作者： Kodati, Dheeraj Tene, Ramakrishnudu Department of Computer Science and Engineering National Institute of Technology Warangal India Department of Computer Science and Engineering Mahindra University Hyderabad India

In recent years, mental health issues have profoundly impacted individuals’ well-being, necessitating prompt identification and intervention. Existing approaches grapple with the complex nature of mental health, facing challenges like task interference, limited adaptability, and difficulty in capturing nuanced linguistic expressions indicative of various conditions. In response to these challenges, our research presents three novel models employing multi-task learning (MTL) to understand mental health behaviors comprehensively. These models encompass soft-parameter sharing-based long short-term memory with attention mechanism (SPS-LSTM-AM), SPS-based bidirectional gated neural networks with self-head attention mechanism (SPS-BiGRU-SAM), and SPS-based bidirectional neural network with multi-head attention mechanism (SPS-BNN-MHAM). Our models address diverse tasks, including detecting disorders such as bipolar disorder, insomnia, obsessive-compulsive disorder, and panic in psychiatric texts, alongside classifying suicide or non-suicide-related texts on social media as auxiliary tasks. Emotion detection in suicide notes, covering emotions of abuse, blame, and sorrow, serves as the main task. We observe significant performance enhancement in the primary task by incorporating auxiliary tasks. Advanced encoder-building techniques, including auto-regressive-based permutation and enhanced permutation language modeling, are recommended for effectively capturing mental health contexts’ subtleties, semantic nuances, and syntactic structures. We present the shared feature extractor called shared auto-regressive for language modeling (S-ARLM) to capture high-level representations that are useful across tasks. Additionally, we recommend soft-parameter sharing (SPS) subtypes-fully sharing, partial sharing, and independent layer-to minimize tight coupling and enhance adaptability. Our models exhibit outstanding performance across various datasets, achieving accuracies of 96.9%, 97.

关键词： Multi-task learning

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Ensemble of AutoEncoder for Restoration of Images Corrupted by Cumulative Combination of Noise 27th

Hierarchical Ensemble of AutoEncoder for Restoration of Imag...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Ganguly, Sayarnil Katham, Sanjana Reddy Agrawal, Sanyam Sinha, Soumen Roy, Rahul Department of Computer Science Engineering Mahindra University Telangana Hyderabad India

ISBN: (纸本)9783031783821

Denoising images corrupted with multiple types of noise is significantly challenging due to their variable noise distribution. These numerous types of noise have a cumulative effect on the image. Therefore, considering only a single or multiple distribution of noise would be ineffective for restoration. Moreover, estimation of the noise is difficult where there is heterogeneity in the combination of the noise. Hence in this paper, we propose a sequence of autoencoders to progressively restore the noisy images. Each autoencoder specializes in the restoration of images affected by a particular combination of noise. These trained autoencoders are arranged in a state-space graph such that all possible combinations of noises can be screened by the autoencoder or their ensembles. The root of the state space graph is the Nth level autoencoder as it is exposed to all kinds of noise. In contrast, the autoencoder at the leaf level is exposed to one type of noise. During restoration of images, the obtained state space tree is searched using a heuristics search algorithm to find the near-optimal ensemble of autoencoders which can denoise the combinations of noises present in an image. Experiments are conducted with diverse combinations of noises to demonstrate the efficacy of the proposed approach. Furthermore, the proposed method is compared with various state-of-the-art approaches. It was observed that the proposed approach had significant efficiency in handling the combination of noise and outperforms various state-of-the-art techniques. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Image denoising

来源：评论

学校读者我要写书评

暂无评论

A self-organized priority-based MAC protocol in wireless sensor networks based on SDR-RHSO optimal relay node selection and HL-ANN wake-up scheduling

引用

Journal of Ambient Intelligence and Humanized Computing 2023年第8期14卷 11093-11102页

作者： Thalagondapati, Veeraiah Singh, Maheshwari Prasad Computer Science and Engineering Mahindra University Telangana Hyderabad India Computer Science and Engineering National Institute of Technology-Patna Bihar Patna India

A budding technology that possesses an extensive range of potential applications comprising environmental monitoring, medical systems, smart spaces, along with robotic exploration is a wireless sensor network (WSN). Medium Access Control (MAC) is a vital technique that evades collisions such that two interfering nodes do not transfer simultaneously for the WSN’s successful operation. However, owing to high energy consumption (EC), non-scalability, Delay, etc., designing an effective MAC protocol for WSN is a complex task. An efficient self-organized MAC protocol in WSN is formed by the work for conquering the issue. Utilizing Linear Scaling centered Social Spider Optimization Algorithm (LS-SSOA), the work is initiated with optimal cluster head (CH) selection under the Mahalanobis Distance centered K-Means Clustering (MD-KMC) technique, which extends the network lifetime (NLT) along with the stability of WSN. Afterward, an optimal relay node (RN) selection is created using Swap Displacement and Reverse-Based Rock Hyraxes Swarm Optimization (SDR-RHSO) for minimizing the energy loss throughout data transmission from sensor node (SN) to CH. Lastly, by identifying the active nodes and non-active nodes using Hebbian Learning-centered Artificial Neural Network (HL-ANN), a wake-up scheduling is performed betwixt the nodes that decrease the extra EC along with delay. A better network lifetime, lower energy consumption, higher throughput value, and accurate wake-up scheduling is attained by the work as exhibited by the experimental analysis and stays effective as analogized to the existing top-notch method. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

关键词： Scheduling

来源：评论

学校读者我要写书评

暂无评论

Multi-Faceted Reinforcement Learning Frameworks for Dynamic Cluster Head Selection and Energy-Efficient Routing in Wireless Sensor Networks

引用

SN computer science 2025年第5期6卷 1-20页

作者： Mouli, S. Sudhamsu Talagondapati, T. Veeraiah Singh, M.P. Department of Computer Science and Engineering Ecole Centrale School of Engineering Mahindra University Hyderabad India Department of Computer Science and Engineering National Institute of Technology Patna India

Wireless Sensor Networks (WSNs) face critical energy efficiency challenges due to resource limitations, especially in extending network lifetime. This paper presents a reinforcement learning-based solution combining Linear Optimization State-Action-Reward-State-Action (LO-SARSA) for energy-aware Cluster Head (CH) selection and a Proportional Optimization Policy Gradient Method (PO-PGM) for optimal routing. CH selection is modeled as an RL problem, where LO-SARSA learns policies based on residual energy and node distance, while PO-PGM uses a neural network to make adaptive routing decisions. Extensive simulations demonstrate that the proposed approach significantly improves energy consumption, network lifetime, and packet delivery ratio, achieving up to a 20% gain in network lifetime over conventional methods. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： CH Selection LO-SARSA Optimal Routing PO-PGM Reinforcement Learning(RL) Wireless Sensor Networks (WSNs)

来源：评论

学校读者我要写书评

暂无评论

Early Detection of Diabetic Retinopathy Using Enhanced Transfer Learning Techniques

Early Detection of Diabetic Retinopathy Using Enhanced Trans...

引用

2024 IEEE International Conference on computer Vision and Machine Intelligence, CVMI 2024

作者： Singh, Navisthi Singh, Navya Mishra, Dipti Mahindra University Department of Artificial Intelligence Hyderabad India Mahindra University Department of Computer Science & Engineering Hyderabad India

ISBN: (纸本)9798350376876

Diabetic Retinopathy (DR) is a critical complication associated with diabetes, often leading to vision loss and blindness if not identified and managed promptly. This project focuses on the early detection and classification of various DR stages, a vital step toward better patient outcomes. We utilize transfer learning with EfficientNetB3, a model recognized for its efficient performance in image classification tasks. Through fine-tuning on a dataset of retinal images and using ImageNet pre-trained weights, the model achieves a higher accuracy in categorizing images into five classes: Healthy, Mild DR, Moderate DR, Proliferative DR, and Severe *** further enhance model robustness, various data augmentation techniques - including rotations, zooms, and horizontal flips - are applied to improve diversity within the training data. The dataset is carefully partitioned into training, validation, and testing subsets, allowing for precise monitoring of the model's generalization capabilities and accuracy across different DR stages. Comprehensive performance evaluation is conducted using metrics such as accuracy, precision, recall, and a confusion matrix to analyze the model's ability to accurately classify each DR stage. This robust approach shows great potential for advancing effective DR classification, ensuring timely and precise treatment strategies for patients at risk. © 2024 IEEE.

关键词： Patient treatment

来源：评论

学校读者我要写书评

暂无评论

PNS Lock: A Portable NUMA-Aware Lock with a Standard Interface 1

PNS Lock: A Portable NUMA-Aware Lock with a Standard Interfa...

引用

1st International Symposium on Parallel Computing and Distributed Systems, PCDS 2024

作者： Gandham, Brahmaiah Alapati, Praveen Ecole Centrale School of Engineering Mahindra University Computer Science and Engineering Hyderabad India

ISBN: (纸本)9798350349658

In shared memory programming, there is a need for synchronization primitives that are suitable for modern hardware to achieve high performance and reduce contention. This paper introduces a novel lock with a standard interface, namely, the PNS Lock. It is a lock for Non-Uniform Memory Access (NUMA) architectures and uses one queue for each NUMA node. PNS lock uses atomic operation for lock handovers between NUMA nodes. This design helps to get better performance by minimizing cross-node communication and contention. PNS lock intelligently adapts to the underlying system topology. This paper discusses the design of a PNS lock and compares it's performance by considering the state-of-the-art synchronization objects. PNS lock also provides a standard interface and ensures seamless integration with existing multi-threaded applications. From the experimental evaluation, it is observed that PNS lock helps to enhance the performance and increases the cache hit rate. This synchronization primitive, PNS lock is a great tool for the developers in NUMA environments. © 2024 IEEE.

关键词： Memory architecture

来源：评论

学校读者我要写书评

暂无评论

Latent Anomalies: Advanced Detection in Industrial Machinery via Convolutional Autoencoder

Latent Anomalies: Advanced Detection in Industrial Machinery...

引用

2024 IEEE International Conference on computer Vision and Machine Intelligence, CVMI 2024

作者： Adithya, B V Soma Siddi, Rohith Bolla, Jasmitha Dupaguntla, Koushik Mishra, Dipti Mahindra University Department of Artificial Intelligence Hyderabad India Mahindra University Department of Computer Science Engineering Hyderabad India

ISBN: (纸本)9798350376876

The paper presents a framework that utilizes Convolutional Autoencoders (CAEs) for anomaly detection in industrial settings. The framework focuses on ensuring product quality and identifying defects in industrial manufacturing, with an emphasis on distinguishing between normal and anomalous occurrences. The proposed model is trained using 25 epochs with a batch size of 1024 and employs Mean Square Error (MSE) and Structural Similarity Index Measure (SSIM) to optimize pixel-wise accuracy and perceptual quality. The Adam optimizer is used to minimize the reconstruction loss, and the model's performance is evaluated on the MVTec Anomaly Detection dataset, achieving an accuracy of 77.1 on the validation set. The study highlights the potential of CAEs in enhancing industrial quality control and emphasizes future work focusing on optimizing model architecture, real-time anomaly detection, and exploring transfer learning techniques for improved versatility and robustness. The experimental results, visualizations, and implications underscore the significance of the proposed framework in addressing the challenges of anomaly detection in industrial manufacturing. © 2024 IEEE.

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

Smart Diagnosis of a Cantilever Beam Using SVM 3

Smart Diagnosis of a Cantilever Beam Using SVM

引用

3rd International Conference on Emerging Techniques in Computational Intelligence, ICETCI 2023

作者： Katam, Rakesh Kalapatapu, Prafulla Pasupuleti, Venkata Dilip Kumar Mahindra University Civil Engineering Hyderabad India Mahindra University Computer Science Engineering Hyderabad India

ISBN: (纸本)9798350300604

Damage assessment (DA) in Structural Health Monitoring (SHM) is receiving a lot of attention from academics, research, and industry due to increased city traffic volumes brought on by urban growth. Using this kind of technique, structures like buildings, bridges, marine structures, etc., may keep track of their own structural integrity both during operation and over the period of their existence. Although reliability increases, the cost of system and structure maintenance and inspection decreases. This kind of structural deterioration testing is performed without actually causing any damage to the system. Before going on to the in-situ experimentation for damage assessment utilizing sensors, a laboratory experiment employing a cantilever steel beam is carried out to understand the modal parameters from the vibration data collected. This study analyzes damage in a cantilever beam using data obtained from vibrations and statistical features derived from time series data. Support Vector Machines (SVMs) are used for both classification and regression tasks. This study achieved a 90% accuracy rate for categorizing the data into damaged and undamaged types using the SVM approach. SVM may be affected by the choice of hyperparameters, such as the regularization parameter and kernel type, and can be computationally expensive for large datasets, but it is still a useful and well-liked approach in the machine-learning community. © 2023 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Prediction Model for Suicidal Behavior Disorder Risk Analysis by Correlating Cyber and Real World Data 4

Prediction Model for Suicidal Behavior Disorder Risk Analysi...

引用

4th IEEE International Conference on Emerging Techniques in Computational Intelligence, ICETCI 2024

作者： Neelakantam, Gone Ecole Centrale School of Engineering Mahindra University Dept. Computer Science and Engineering Hyderabad India

ISBN: (纸本)9798350387063

In the era of internet of things (IoT), people are attentive to express their non verbal behavior on cyber world such as social media by using smart phone, laptop or tablet. People are more active in sharing their daily activities status such as current activity the person is into and personal life situations such as achievement, problem, stress and even any hazardous desire (i.e. suicide). In contrast with real world such as smart home environment, people generally spent their non verbal behavior in term of Activity Daily Livings (ADLs) in associating with smart home sensors. Based on non verbal and verbal behaviors, data are generated from both Cyber and Real world, which are big in volume and variety. However, there is lack of investigating Cyber and Real world data especially in analyzing the risk of committing suicide, considering suicide is a big issues and threats in the society. Therefore, it motivate us to propose prediction model to determine high and low risk of committing suicide by combining two separate approaches such as Activity Recognition (AR) for real world and Sentiment Analysis (SA) for cyber world. © 2024 IEEE.

关键词： Risk assessment

来源：评论

学校读者我要写书评

暂无评论

Optimizing Wireless Charging Schedules in Sensor Networks: A Hybrid GSA-SA Approach for On-Demand Mobile Charging 14

Optimizing Wireless Charging Schedules in Sensor Networks: A...

引用

14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023

作者： Mouli, Seethamraju Sudhamsu Talagondapathi, Veeraiah Kongara, Mahesh Chowdary Mahindra University Dept of Computer Science and Engineering Hyderabad India

ISBN: (纸本)9798350335095

Prior work on Wireless Sensor Networks (WSNs) has found that Sensor Node (SN) low energy capacity limits their ongoing operation. Recent discoveries in the field of Wireless Energy Transfer (WET) technology has emerged as a viable solution to sensor nodes' low energy capacity. It creates dependable Wireless Rechargeable Sensor Networks (WRSNs). Recently, there has been a lot of interest in including Wireless Mobile Chargers (WMCs) in WRSNs to refill the energy of key sensor nodes (SNs) through WET. The basic concept of WET is the use of one or more (WMCs) to recharge SNs, resulting in the development of (WRSNs). In on-demand WRSN, optimize the order in which the SNs are charged (i.e., the charging schedule). Furthermore, when creating a WMC's charging schedule, we need to consider that requesting SNs creates both space and time constraints by considering these parameters. In this paper, we propose a hybrid Gravitational Search Algorithm-Simulated Annealing (GSA-SA) based recharge schedule for replenishing the sensor nodes which can be used to optimize sensor node deployment to maximize network coverage, reduce energy usage, and increase network lifespan. Our solution includes a unique strategy for efficient fitness functions. We do extensive simulations on the proposed system to show how it outperforms novel algorithms like First Come, First Served (FCFS), Nearest Job Next with Preemption (NJNP), and Gravitational Search Algorithm (GSA). In terms of charging latency. © 2023 IEEE.

关键词： Sensor nodes

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：