LLM-based agents have emerged as transformative tools capable of executing complex tasks through iterative planning and action, achieving significant advancements in understanding and addressing user needs. Yet, their effectiveness remains limited in specialized domains such as mental health diagnosis, where they underperform compared to general applications. Current approaches to integrating diagnostic capabilities into LLMs rely on scarce, highly sensitive mental health datasets, which are challenging to acquire. These methods also fail to emulate clinicians' proactive inquiry skills, lack multi-turn conversational comprehension, and struggle to align outputs with expert clinical reasoning. To address these gaps, we propose DSM5AgentFlow, the first LLM-based agent workflow designed to autonomously generate DSM-5 Level-1 diagnostic questionnaires. By simulating therapist-client dialogues with specific client profiles, the framework delivers transparent, step-by-step disorder predictions, producing explainable and trustworthy results. This workflow serves as a complementary tool for mental health diagnosis, ensuring adherence to ethical and legal standards. Through comprehensive experiments, we evaluate leading LLMs across three critical dimensions: conversational realism, diagnostic accuracy, and explainability. Our datasets and implementations are fully open-sourced.

ABSTRACT With the proliferation of cloud computing and the escalating demand for extensive data processing capabilities, an increasing number of enterprises are embracing hybrid cloud solutions. However, as more businesses move toward hybrid clouds, the need for effective solutions to privacy and security concerns becomes increasingly important. Although current scheduling approaches for cloud computing have addressed privacy protection to some extent, few have adequately considered the unique challenges posed by hybrid clouds. To address this gap, we propose a novel approach for scheduling jobs in hybrid clouds that prioritizes privacy protection. Our approach, called PH‐DRL, leverages Deep Reinforcement Learning (DRL) to intelligently allocate jobs to virtual machines, optimizing both privacy and Quality of Service (QoS), while minimizing response time. We present the detailed implementation of our approach and our experimental results demonstrate the superior performance of PH‐DRL in terms of privacy protection compared to existing methods.

ABSTRACT Polyhedral compilers and tensor compilers have achieved great success on accelerating scientific computing kernels and deep learning networks, respectively. Although much work has been done to integrate techniques of the polyhedral model to tensor compilers for accelerating deep learning, leveraging the powerful auto‐tuning ability of modern tensor compilers to accelerate more general scientific computing kernels is challenging and is still at its dawn. In this work, we introduce a method to accelerate a family of basic scientific computing kernels by fusing the polyhedral compiler Pluto and the tensor compiler Tensor Virtual Machine (TVM) to generate efficient implementations targeting the heterogeneous CPU/GPU platform. The fusion is done in four steps: building a polyhedral model for the loop description of a given scientific kernel; designing schedules to transform the polyhedral model to new ones to enable rectangular tiling and expose explicit parallelism; selecting a new polyhedral model and converting it to the tensor compute representation; auto‐tuning the tensor compute to generate efficient implementations on both CPUs and GPUs. Shifting and padding optimizations are also considered to avoid conditionals. Experiments on 30 typical scientific computing kernels show that our method achieves speedup on average over a typical polyhedral compiler PPCG on GPU.

Scheduling workflows in hybrid cloud environments presents significant challenges due to the inherent complexity of workflows and the dynamic nature of cloud resources. This complexity is further increased when attempting to balance workflow performance with privacy protection. Recent efforts have leveraged deep reinforcement learning (DRL) to address these challenges. However, most of these approaches rely on single-agent models, which can lead to security issues and scalability problems due to their centralized processing. Specifically, the properties of workflows are transferred to the single agent, which risks leaking privacy information. Our paper addresses these issues by introducing MARS, a real-time workflow scheduling method that prioritizes privacy protection in hybrid clouds. MARS leverages multi-agent deep reinforcement learning (MADRL) to optimize the workflow scheduling of cloud virtual machines (VMs). The benefit of our solution is that it relies on the collaborative learning of multi-agents on multiple VMs, which could assign user data to specific cloud servers for privacy protection while sharing training experiences between agents. In our implementation, MARS aims to reduce workflow completion time and operational costs while complying with strict privacy protection guidelines. The experimental results demonstrate that MARS can significantly surpass existing methods, reducing makespan by an average of $53.18 \%$ and costs by $61.98 \%$ compared to basic techniques, and achieving $20.26 \%$ and $25.71 \%$ improvements over the latest advanced methods, respectively.

With the wide application of Internet of Things (IoT) technology, large volumes of multimodal data are collected and analyzed for various diagnoses, analyses, and predictions to help in decision-making and management. However, the research on protecting data integrity and privacy is quite limited, while the lack of proper protection for sensitive data may have significant impacts on the benefits and gains of data owners. In this research, we propose a protection solution for data integrity and privacy. Specifically, our system protects data integrity through distributed systems and blockchain technology. Meanwhile, our system guarantees data privacy using differential privacy and Machine Learning (ML) techniques. Our system aims to maintain the usability of the data for further data analytical tasks of data users, while encrypting the data according to the requirements of data owners. We implement our solution with smart contracts, distributed file systems, and ML models. The experimental results show that our proposed solution can effectively encrypt source IoT data according to the requirements of data users while data integrity can be protected under the blockchain.

Predictive Process Monitoringaims to predict the future information of ongoing process executions by leveraging machine and deep learning techniques. One of the tasks is known asremaining time prediction, which focuses on predicting the remaining time of ongoing cases. Accurate remaining time prediction can be valuable and important for improving business operations or taking timely interventions to prevent delays. For predicting the remaining time, existing work has used deep learning techniques to achieve high prediction accuracy. However, most of these techniques tend to learn very complex models that are difficult to explain. Systematic feature selection approaches may help improve both the prediction accuracy and the explainability of the model. In this paper, we introduce a feature-informed cascade prediction framework to predict the remaining time. Specifically, we first propose an approach that builds a tree of features by systematically estimating their effects on the remaining time prediction. Next, we use the tree to either automatically select an optimal combination of features or to guide users in this selection process. Each selected feature is correlated with its prediction results in our Feature-informed Cascade Prediction Model (FCPM) for explainability. The proposed approach has been implemented and is made publicly available. Using eight public real-life event logs, the proposed approach is compared to the state-of-the-art approaches in terms of prediction accuracy. In addition, it is demonstrated that our approach visualizes the impact of each input feature in the prediction of individual cases, producing explanations of the prediction results.

The amalgamation of artificial intelligence with Internet of Things (AIoT) devices have seen a rapid surge in growth, largely due to the effective implementation of deep neural network (DNN) models across various domains. However, the deployment of DNNs on such devices comes with its own set of challenges, primarily related to computational capacity, storage, and energy efficiency. This survey offers an exhaustive review of techniques designed to accelerate DNN inference on AIoT devices, addressing these challenges head-on. We delve into critical model compression techniques designed to adapt to the limitations of devices and hardware optimization strategies that aim to boost efficiency. Furthermore, we examine parallelization methods that leverage parallel computing for swift inference, as well as novel optimization strategies that fine-tune the execution process. This survey also casts a future-forward glance at emerging trends, including advancements in mobile hardware, the co-design of software and hardware, privacy and security considerations, and DNN inference on AIoT devices with constrained resources. All in all, this survey aspires to serve as a holistic guide to advancements in the acceleration of DNN inference on AIoT devices, aiming to provide sustainable computing for upcoming IoT applications driven by artificial intelligence.

With the prevalence of big-data-driven applications, such as face recognition on smartphones and tailored recommendations from Google Ads, we are on the road to a lifestyle with significantly more intelligence than ever before. Various neural network powered models are running at the back end of their intelligence to enable quick responses to users. Supporting those models requires lots of cloud-based computational resources, e.g., CPUs and GPUs. The cloud providers charge their clients by the amount of resources that they occupy. Clients have to balance the budget and quality of experiences (e.g., response time). The budget leans on individual business owners, and the required Quality of Experience (QoE) depends on usage scenarios of different applications. For instance, an autonomous vehicle requires an real-time response, but unlocking your smartphone can tolerate delays. However, cloud providers fail to offer a QoE-based option to their clients. In this paper, we proposeDQoES, differentiated quality of experience scheduler for deep learning inferences.DQoESaccepts clients’ specifications on targeted QoEs, and dynamically adjusts resources to approach their targets. Through the extensive cloud-based experiments,DQoESdemonstrates that it can schedule multiple concurrent jobs with respect to various QoEs and achieve up to 8x times more satisfied models when compared to the existing system.

Businesses have made increasing adoption and incorporation of cloud technology into internal processes in the last decade. The cloud-based deployment provides on-demand availability without active management. More recently, the concept of cloud-native application has been proposed and represents an invaluable step toward helping organizations develop software faster and update it more frequently to achieve dramatic business outcomes. Cloud-native is an approach to build and run applications that exploit the cloud computing delivery model’s advantages. It is more about how applications are created and deployed than where. The container-based virtualization technology, e.g., Docker and Kubernetes, serves as the foundation for cloud-native applications. This paper evaluates the performance of deep learning applications in a cloud-native environment.

As a part of Advanced Driver Assistance Systems (ADASs), Consensus-based Speed Advisory Systems (CSAS) have been proposed to recommend a common speed to a group of vehicles for specific application purposes, such as emission control and energy management. With Vehicle-to-Vehicle (V2V), Vehicle-to-Infrastructure (V2I) technologies and advanced control theories in place, state-of-the-art CSAS can be designed to get an optimal speed in a privacy-preserving and decentralized manner. However, the current method only works for specific cost functions of vehicles, and its execution usually involves many algorithm iterations leading long convergence time. Therefore, the state-of-the-art design method is not applicable to a CSAS design which requires real-time decision making. In this article, we address the problem by introducing MPC-CSAS, a Multi-Party Computation (MPC) based design approach for privacy-preserving CSAS. Our proposed method is simple to implement and applicable to all types of cost functions of vehicles. Moreover, our simulation results show that the proposed MPC-CSAS can achieve very promising system performance in just one algorithm iteration without using extra infrastructure for a typical CSAS.

Load balancing is directly associated with the overall performance of a parallel and distributed computing system. Although the relevant problems in communication and computation have been well studied in data center environments, few works have considered the issues in an Internet of Things (IoT) edge scenario. In fact, processing data in a load balancing way for the latter case is more challenging. The main reason is that, unlike a data center, both the data sources and the network infrastructure in an IoT edge system can be dynamic. Moreover, with different performance requirements from IoT networks and edge servers, it will be hard to characterize the performance model and to perform runtime optimization for the whole system. To tackle this problem, in this work, we propose a load-balancing aware networking approach for efficient data processing in IoT edge systems. Specifically, we introduce an IoT network dynamic clustering solution using the emerging deep reinforcement learning (DRL), which can both fulfill the communication balancing requirements from IoT networks and the computation balancing requirements from edge servers. Moreover, we implement our system with a long short term memory (LSTM) based Dueling Double Deep Q-Learning Network (D3QN) model, and our experiments with real-world datasets collected from an autopilot vehicle demonstrate that our proposed method can achieve significant performance improvement compared to benchmark solutions.

MOTIVATION: The analysis of gene co-expression network (GCN) is critical in examining the gene-gene interactions and learning the underlying complex yet highly organized gene regulatory mechanisms. Numerous clustering methods have been developed to detect communities of co-expressed genes in the large network. The assumed independent community structure, however, can be oversimplified and may not adequately characterize the complex biological processes. RESULTS: We develop a new computational package to extract interconnected communities from gene co-expression network. We consider a pair of communities be interconnected if a subset of genes from one community is correlated with a subset of genes from another community. The interconnected community structure is more flexible and provides a better fit to the empirical co-expression matrix. To overcome the computational challenges, we develop efficient algorithms by leveraging advanced graph norm shrinkage approach. We validate and show the advantage of our method by extensive simulation studies. We then apply our interconnected community detection method to an RNA-seq data from The Cancer Genome Atlas (TCGA) Acute Myeloid Leukemia (AML) study and identify essential interacting biological pathways related to the immune evasion mechanism of tumor cells. AVAILABILITY: The software is available at Github: https://github.com/qwu1221/ICN and Figshare: https://figshare.com/articles/software/ICN-package/13229093. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Software-defined networking (SDN) is a cornerstone of next-generation networks and has already led to numerous advantages for data-center networks and wide-area networks. However, SDN is not widely adopted in constrained networks, such as Wireless Sensor Networks (WSN), due to excessive control overhead, lossy medium, and in-band control channels. Therefore, a key challenge to enable Software-Defined Wireless Sensor Networks (SD-WSN) is to reduce the number of control messages required to configure the data plane. In this paper, we propose a cluster-based flow control approach in hybrid SDNs. Our approach is hybrid in the sense that it takes advantage of distributed legacy routing and centralized SDN routing. In addition, it makes a trade-off between the granularity of flow control and the communication overhead induced by the SDN controller. The approach partitions a network into clusters with minimum number of border nodes. Instead of handling the individual flows of each node, the SDN controller only manages incoming and outgoing traffic flows of clusters through border nodes, while the flows inside each cluster are controlled by a distributed legacy WSN routing algorithm. Our proof-of-concept implementations in both software and hardware show that our approach is efficient with respect to reducing the number of nodes that must be managed and the number of control messages. In comparison to benchmark solutions with and without clustering, our solution reduces communication costs for flow configuration in an SD-WSN at least by 27% and at most by 88% respectively, without degrading packet delay nor delivery rate.

Large data centers are currently the mainstream infrastructures for big data processing. As one of the most fundamental tasks in these environments, the efficient execution of distributed data operators (e.g., join and aggregation) are still challenging current data systems, and one of the key performance issues is network communication time. State-of-the-art methods trying to improve that problem focus on either application-layer data locality optimization to reduce network traffic or on network-layer data flow optimization to increase bandwidth utilization. However, the techniques in the two layers are totally independent from each other, and performance gains from a joint optimization perspective have not yet been explored. In this article, we propose a novel approach called NEAL (NEtwork-Aware Locality scheduling) to bridge this gap, and consequently to further reduce communication time for distributed big data operators. We present the detailed design and implementation of NEAL, and our experimental results demonstrate that NEAL always performs better than current approaches for different workloads and network bandwidth configurations.

Processing big data generated in large Internet of Things (IoT) networks is challenging current techniques. To date, a lot of network clustering approaches have been proposed to improve the performance of data collection in IoT. However, most of them focus on partitioning networks with static topologies, and thus they are not optimal in handling the case with moving objects in the networks. Moreover, to the best of our knowledge, none of them has ever considered the performance of computing in edge servers. To solve these problems, we propose a highly efficient IoT network dynamic clustering solution in edge computing using deep reinforcement learning (DRL). Our approach can both fulfill the data communication requirements from IoT networks and load-balancing requirements from edge servers, and thus provide a great opportunity for future high performance IoT data analytics. We implement our approach using a Deep Q-learning Network (DQN) model, and our preliminary experimental results show that the DQN solution can achieve higher scores in cluster partitioning compared with the current static benchmark solution.

As deployment of Internet of Things (IoT) devices gain momentum, there is an increased interest in implementing machine learning (ML) algorithms on IoT devices. Most of the existing ML solutions, however, rely on a central server to execute data-intensive ML models, because most devices in IoT systems do not have sufficient storage and computing resources. This paper presents a distributed Artificial Neural Networks (ANN) architecture, called PCANN, which allows execution of a complex image recognition task on a collection of resource-constrained IoT devices. Our solution separates a single ML model into multiple small modules that are executed by the distributed IoT devices. The solution effectively reduces storage and computing requirements for individual devices to store and process ML model. We design multiple PCANN models and utilize the models for human posture recognition as the case study. The experimental results show that the distributed PCANN architecture achieves comparable accuracy as the classical ANN model, while the average size of each PCANN module is largely reduced.

Thread is a wireless networking protocol based on existing open standards such as IEEE 802.15.4, 6LoWPAN, and IPv6, and is designed for low power, low data rate and short-range applications. In this paper, we evaluate the performance of Thread protocol for lighting systems using a practical testbed setup in an office area. We focus our evaluation on unicast and multicast, which are the two most important communication paradigms for lighting applications. The unicast and multicast of Thread are tested under various communication parameters, such as concurrent multicast sources, hop distance from the source to the destination, communication interval, etc. The key performance indicators include end-to-end latency, time for complete coverage, synchronization, and packet delivery ratio. In addition, we use the measurement results of unicast and multicast in some typical lighting applications to analyze its impact on user experience. Through our extensive experiment and analysis, we conclude that the performance of Thread multicast and unicast has much leeway to meet the requirements of lighting systems.

Process models as knowledge graph representation have been widely used in various domains to create products and deliver services. Although different process model discovery approaches have been proposed in recent years, few of them are designed for distributed computing environments. Specifically, none of them has been studied in the emerging edge computing application scenarios. In this paper, based on the requirements of some real-time process services, we propose a system design for learning process models in IoT edge. We present the details of our solution and our preliminary results on a simulated IoT network show that our method can discover real-time process models in less than a second.

Software-defined networking (SDN) is a cornerstone of next-generation networks and has already led to numerous advantages for data-center networks and wide-area networks, for instance in terms of reduced management complexity and more fine-grained traffic engineering. However, the design and implementation of SDN within wireless sensor networks (WSN) have received far less attention. Unfortunately, because of the multi-hop type of communication in WSN, a direct reuse of the wired SDN architecture could lead to excessive communication overhead. In this paper, we propose a cluster-based flow management approach that makes a trade-off between the granularity of monitoring by an SDN controller and the communication overhead of flow management. A network is partitioned into clusters with a minimum number of border nodes. Instead of having to handle the individual flows of all nodes, the SDN controller only manages incoming and outgoing traffic flows of clusters through border nodes. Our proof-of-concept implementations in software and hardware show that, when compared with benchmark solutions, our approach is significantly more efficient with respect to the number of nodes that must be managed and the number of control messages exchanged.

A wireless power transfer network (WPTN) aims to support devices with cable-less energy on-demand. Unfortunately, wireless power transfer itself-especially through radio frequency radiation rectification-is fairly inefficient due to decaying power with distance, antenna polarization, etc. Consequently, idle charging needs to be minimized to reduce the already large costs of providing energy to the receivers. In turn, energy saving in a WPTN can be boosted by simply switching off the energy transmitter when the received energy is too weak for rectification. Therefore in this paper we propose, and experimentally evaluate, two “green” protocols for the control plane of static charger/mobile receiver WPTN aimed at optimizing the charger workflow to make the WPTN reduce idle time of transmitters. Those protocols are: “beaconing,” where receivers advertise their presence to the WPTN, and “probing” exploiting the receiver feedback from the WPTN on the level of received energy. We demonstrate that both protocols reduce the unnecessary WPTN uptime, however trading it for the reduced energy provision, compared to the base case of “WPTN charger always on.” For example, our system (in our experiments) saves at most ≈80 % of energy at the charger with only ≈17% less energy possibly harvested.

Today's wireless networks are connecting more and more devices around us, leading to the birth of a new distributed computing platform, in the form of a spatial computer. The main difference with traditional computing models is that space and time become intertwined with computation, especially when scaling up the system. Computations performed by each element are now related to its spatial position. This property is the key ingredient when assuring the availability for various distributed networking services and applications. Computations become linked to the concept of space. Estimating distances between components (especially in dynamic networks characterized by the node mobility) thus becomes one of the most important building blocks for spatial computing. The majority of the algorithms that come from the MANET community presume knowledge about node position via systems such as GPS, or employ a one-time manual network topology configuration. While for some application scenarios this approach is feasible, for a lot of cases it suffers from frequent unavailability (e.g. indoors) and high costs in terms of energy consumption. Therefore, intense demand exists for a new kind of distance estimation algorithm using only simple local interactions, without knowledge of global information. The main contribution of the article is the introduction of a novel distributed algorithm, called gradient-based distance estimation (GDE), for the estimation of distances in networks characterized by mobility, specifically targeting the context of spatial computing. GDE is based on a gossiping mechanism to estimate distances between nodes with only local interactions. It significantly improves current state of the art by employing statistical analysis and making better use of the information available at each node.We analyze the parameters that should be considered by real applications, and present mathematical models to compensate their influence for distance estimation. Three spatial computing applications using GDE are presented: geographical cluster center detection, topological overlay shape construction and geographic routing. The simulation-based evaluation shows that GDE succeeds in estimating the distance between nodes in both static and mobile scenarios with considerably high accuracy for various simulations setups, such as varying node density, node speed or spatial node distribution.

Today, wireless networks are connecting most of the devices around us. The scale of these systems demands for novel techniques to maintain availability for various services such as routing, localization, context detection etc. Distance estimation is one of their most important building blocks. The majority of current algorithms, presumes knowledge about node position via systems such as GPS. While for some application scenarios this approach is feasible, for a lot of cases it suffers from frequent unavailability and high costs in terms of energy consumption. The main contribution of this paper is the introduction of a novel distributed algorithm called GDE, for the estimation of distances in large-scale wireless networks. GDE is a mechanism which estimates distances between nodes based solely on local interactions. The evaluation by means of simulations shows that GDE succeeds in estimating the distance between nodes in both static and mobile scenarios with considerably high accuracy, even under the influence of different kinds of environment parameters, such as node density, node speed, spatial node distribution, multicast percentage, etc.