AC optimal power flow (AC OPF) is a fundamental problem in power system operations. Accurately modeling the network physics via the AC power flow equations makes AC OPF a challenging nonconvex problem. To search for g...
详细信息
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, faci...
详细信息
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.
Realizing digital-twin services is one of promising applications in 6 G mobile communication and network scenarios. In addition, the use of unmanned aerial vehicles (UAVs) is essential for enabling the services e...
详细信息
Realizing digital-twin services is one of promising applications in 6 G mobile communication and network scenarios. In addition, the use of unmanned aerial vehicles (UAVs) is essential for enabling the services even in the extreme areas where humans cannot reach. In this emerging scenario, it is necessary to design collaborative algorithms for autonomous UAV trajectory control and a centralized computing platform (e.g., cloud) in digital-twin networks. For this system, it is required to build energy-efficient algorithms due to the power-hungry nature in UAVs. Based on this requirements and system characteristics, this paper proposes autonomous UAV charging algorithms and systems where the UAVs are classified into two types, i.e., cluster UAVs (for main image recording operations in digital-twin services, and some of them take the roles of mobile edge computing) and charging UAVs (for charging the cluster UAVs). Our proposed charging should be (i) fully distributed for practical, scalable, and low-overhead operations and (ii) trustworthy for secure and privacy-preserving computation;where these are essential for collaborative operations. Therefore, a novel auction-based charging algorithm for UAV-based digital-twin networks is proposed in order to realize the distributed and truthful operations, which cannot be achieved by the convex optimization-based centralized algorithms in the literature. Our performance evaluation verifies that the proposed algorithm achieves performance improvements (at most 15.53%). IEEE
The process of modifying digital images has been made significantly easier by the availability of several image editing software. However, in a variety of contexts, including journalism, judicial processes, and histor...
详细信息
The need for renewable energy access has led to the use of variable input converter approaches because renewable energy sources often generate electricity in an unpredictable manner. A high-performance multi-input boo...
详细信息
The need for renewable energy access has led to the use of variable input converter approaches because renewable energy sources often generate electricity in an unpredictable manner. A high-performance multi-input boost converter is developed to provide the necessary output voltage and power while accommodating variations in input sources. This converter is specifically designed for the efficient usage of renewable energy. The proposed architecture integrates three separate unidirectional input power sources: photovoltaics, fuel cells, and storage system batteries. The architecture has five switches, and the implementation of each switch in the converter is achieved by applying the calculated duty ratios in various operating states. The closed-loop response of the converter with a proportional-integral (PI) controller-based switching system is examined by analyzing the Matlab-Simulink model utilizing a proportional-integral derivative (PID) tuner. The controller can deliver the desired output voltage of 400 V and an average power of 2 kW while exhibiting low switching transient effects. Therefore, the proposed multi-input interleaved boost converter demonstrates robust results for real-time applications by effectively harnessing renewable power sources.
Exploration strategy design is a challenging problem in reinforcement learning(RL),especially when the environment contains a large state space or sparse *** exploration,the agent tries to discover unexplored(novel)ar...
详细信息
Exploration strategy design is a challenging problem in reinforcement learning(RL),especially when the environment contains a large state space or sparse *** exploration,the agent tries to discover unexplored(novel)areas or high reward(quality)*** existing methods perform exploration by only utilizing the novelty of *** novelty and quality in the neighboring area of the current state have not been well utilized to simultaneously guide the agent’s *** address this problem,this paper proposes a novel RL framework,called clustered reinforcement learning(CRL),for efficient exploration in *** adopts clustering to divide the collected states into several clusters,based on which a bonus reward reflecting both novelty and quality in the neighboring area(cluster)of the current state is given to the *** leverages these bonus rewards to guide the agent to perform efficient ***,CRL can be combined with existing exploration strategies to improve their performance,as the bonus rewards employed by these existing exploration strategies solely capture the novelty of *** on four continuous control tasks and six hard-exploration Atari-2600 games show that our method can outperform other state-of-the-art methods to achieve the best performance.
This paper presents a comprehensive dataset of LoRaWAN technology path loss measurements collected in an indoor office environment, focusing on quantifying the effects of environmental factors on signal propagation. U...
详细信息
This paper presents a novel supervised learning framework for real-time optimization of multi-parametric mixed-integer quadratic programming (mp-MIQP) problems. The framework utilizes a multi-layer perceptron (MLP) mo...
详细信息
Federated learning is widely accepted as a privacy-preserving paradigm for training a shared global model across multiple client devices in a collaborative fashion. However, in practice, the significantly limited comp...
详细信息
Federated learning is widely accepted as a privacy-preserving paradigm for training a shared global model across multiple client devices in a collaborative fashion. However, in practice, the significantly limited computational power on client devices has been a major barrier when we wish to train large models with potentially hundreds of millions of parameters. In this paper, we propose a new architecture, referred to as Infocomm, that incorporates locally supervised learning in federated learning. With locally supervised learning, the disadvantages of split learning can be avoided by using a more flexible way to offload training from resource constrained clients to a more capable server. Infocomm enables parallel training of different modules of the neural network in both the server and clients in a gradient-isolated fashion. The efficacy in reducing both training time and communication time is supported by our theoretical analysis and empirical results. In the scenario involving larger models and fewer available local data, Infocomm has been observed to reduce the elapsed time per round by over 37% without sacrificing accuracy compared to both conventional federated learning or directly combining federated learning and split learning, which showcases the advantages of Infocomm under power-constrained IoT scenarios. IEEE
We consider a power system whose electric demand pertaining to freshwater production is high(high freshwater electric demand),as in the Middle East,and investigate the tradeoff of storing freshwater in tanks versus st...
详细信息
We consider a power system whose electric demand pertaining to freshwater production is high(high freshwater electric demand),as in the Middle East,and investigate the tradeoff of storing freshwater in tanks versus storing electricity in batteries at the day-ahead operation *** storing freshwater and storing electricity increase the actual electric demand at valley hours and decrease it at peak hours,which is generally beneficial in term of cost and ***,to what extent?We analyze this question considering three power systems with different generation-mix configurations,i.e.,a thermal-dominated mix,a renewable-dominated one,and a fully renewable *** generation-mix configurations are inspired by how power systems may evolve in different countries in the Middle *** production uncertainty is compactly modeled using chance *** draw conclusions on how both storage facilities(freshwater and electricity)complement each other to render an optimal operation of the power system.
暂无评论