Nguyễn Hà Trân - Academia.edu (original) (raw)
Papers by Nguyễn Hà Trân
IEEE Transactions on Parallel and Distributed Systems
There is growing interest in applying distributed machine learning to edge computing, forming fed... more There is growing interest in applying distributed machine learning to edge computing, forming federated edge learning. Federated edge learning faces non-i.i.d. and heterogeneous data, and the communication between edge workers, possibly through distant locations and with unstable wireless networks, is more costly than their local computational overhead. In this work, we propose DONE, a distributed approximate Newton-type algorithm with fast convergence rate for communication-efficient federated edge learning. First, with strongly convex and smooth loss functions, DONE approximates the Newton direction in a distributed manner using the classical Richardson iteration on each edge worker. Second, we prove that DONE has linear-quadratic convergence and analyze its communication complexities. Finally, the experimental results with non-i.i.d. and heterogeneous data show that DONE attains a comparable performance to the Newton's method. Notably, DONE requires fewer communication iterations compared to distributed gradient descent and outperforms DANE and FEDL, state-of-the-art approaches, in the case of non-quadratic loss functions.
arXiv (Cornell University), Jun 15, 2020
Federated learning (FL) is a decentralized and privacy-preserving machine learning technique in w... more Federated learning (FL) is a decentralized and privacy-preserving machine learning technique in which a group of clients collaborate with a server to learn a global model without sharing clients' data. One challenge associated with FL is statistical diversity among clients, which restricts the global model from delivering good performance on each client's task. To address this, we propose an algorithm for personalized FL (pFedMe) using Moreau envelopes as clients' regularized loss functions, which help decouple personalized model optimization from the global model learning in a bi-level problem stylized for personalized FL. Theoretically, we show that pFedMe's convergence rate is state-of-the-art: achieving quadratic speedup for strongly convex and sublinear speedup of order 2/3 for smooth nonconvex objectives. Experimentally, we verify that pFedMe excels at empirical performance compared with the vanilla FedAvg and Per-FedAvg, a meta-learning based personalized FL algorithm.
IEEE Internet of Things Journal
Non-terrestrial networks (NTNs), which integrate space and aerial networks with terrestrial syste... more Non-terrestrial networks (NTNs), which integrate space and aerial networks with terrestrial systems, are a key area in the emerging sixth-generation (6G) wireless networks. As part of 6G, NTNs must provide pervasive connectivity to a wide range of devices, including smartphones, vehicles, sensors, robots, and maritime users. However, due to the high mobility and deployment of NTNs, managing the space-air-sea (SAS) NTN resources, i.e., energy, power, and channel allocation, is a major challenge. The design of a SAS-NTN for energy-efficient resource allocation is investigated in this study. The goal is to maximize system energy efficiency (EE) by collaboratively optimizing user equipment (UE) association, power control, and unmanned aerial vehicle (UAV) deployment. Given the limited payloads of UAVs, this work focuses on minimizing the total energy cost of UAVs (trajectory and transmission) while meeting EE requirements. A mixed-integer nonlinear programming problem is proposed, followed by the development of an algorithm to decompose, and solve each problem distributedly. The binary (UE association) and continuous (power, deployment) variables are separated using the Bender decomposition (BD), and then the Dinkelbach algorithm (DA) is used to convert fractional programming into an equivalent solvable form in the subproblem. A standard optimization solver is utilized to deal with the complexity of the master problem for binary variables. The alternating direction method of multipliers (ADMM) algorithm is used to solve the subproblem for the continuous variables. Our proposed algorithm provides a suboptimal solution, and simulation results demonstrate that the proposed algorithm achieves better EE than baselines.
arXiv (Cornell University), Apr 27, 2020
In recent years, the emerging topics of recommender systems that take advantage of natural langua... more In recent years, the emerging topics of recommender systems that take advantage of natural language processing techniques have attracted much attention, and one of their applications is the Conversational Recommender System (CRS). Unlike traditional recommender systems with contentbased and collaborative filtering approaches, CRS learns and models user's preferences through interactive dialogue conversations. In this work, we provide a summarization on the recent evolution of CRS, where deep learning approaches are applied to CRS and have produced fruitful results. We first analyze the research problems and present key challenges in the development of Deep Conversational Recommender Systems (DCRS), then present the current state of the field taken from the most recent researches, including the most common deep learning models that benefit DCRS. Finally, we discuss future directions of this vibrant area.
World Wide Web
In modern days, making recommendation for news articles poses a great challenge due to vast amoun... more In modern days, making recommendation for news articles poses a great challenge due to vast amount of online information. However, providing personalized recommendations from news articles, which are the sources of condense textual information is not a trivial task. A recommendation system needs to understand both the textual information of a news article, and the user contexts in terms of long-term and temporary preferences via the user’s historic records. Unfortunately, many existing methods do not possess the capability to meet such need. In this work, we propose a neural deep news recommendation model called CupMar, that not only is able to learn the user-profile representation in different contexts, but also is able to leverage the multi-aspects properties of a news article to provide accurate, personalized news recommendations to users. The main components of our CupMar approach include the News Encoder and the User-Profile Encoder. Specifically, the News Encoder uses multiple...
IEEE Transactions on Neural Networks and Learning Systems
Emerging cross-device artificial intelligence (AI) applications require a transition from convent... more Emerging cross-device artificial intelligence (AI) applications require a transition from conventional centralized learning systems toward large-scale distributed AI systems that can collaboratively perform complex learning tasks. In this regard, democratized learning (Dem-AI) lays out a holistic philosophy with underlying principles for building large-scale distributed and democratized machine learning systems. The outlined principles are meant to study a generalization in distributed learning systems that go beyond existing mechanisms such as federated learning (FL). Moreover, such learning systems rely on hierarchical self-organization of well-connected distributed learning agents who have limited and highly personalized data and can evolve and regulate themselves based on the underlying duality of specialized and generalized processes. Inspired by Dem-AI philosophy, a novel distributed learning approach is proposed in this article. The approach consists of a self-organizing hierarchical structuring mechanism based on agglomerative clustering, hierarchical generalization, and corresponding learning mechanism. Subsequently, hierarchical generalized learning problems in recursive forms are formulated and shown to be approximately solved using the solutions of distributed personalized learning problems and hierarchical Manuscript
IEEE Transactions on Communications, 2021
Unmanned aerial vehicles (UAVs) can provide an effective solution for improving the coverage, cap... more Unmanned aerial vehicles (UAVs) can provide an effective solution for improving the coverage, capacity, and the overall performance of terrestrial wireless cellular networks. In particular, UAV-assisted cellular networks can meet the stringent performance requirements of the fifth generation new radio (5G NR) applications. In this paper, the problem of energy-efficient resource allocation in UAV-assisted cellular networks is studied under the reliability and latency constraints of 5G NR applications. The framework of ruin theory is employed to allow solar-powered UAVs to capture the dynamics of harvested and consumed energies. First, the surplus power of every UAV is modeled, and then it is used to compute the probability of ruin of the UAVs. The probability of ruin denotes the vulnerability of draining out the power of a UAV. Next, the probability of ruin is used for efficient user association with each UAV. Then, power allocation for 5G NR applications is performed to maximize the achievable network rate using the water-filling approach. Simulation results demonstrate that the proposed ruin-based scheme can enhance the flight duration up to 61% and the number of served users in a UAV flight by up to 58%, compared to a baseline SINR-based scheme.
2019 IEEE Global Communications Conference (GLOBECOM), 2019
Federated learning (FL) rests on the notion of training a global model in a decentralized manner.... more Federated learning (FL) rests on the notion of training a global model in a decentralized manner. Under this setting, mobile devices perform computations on their local data before uploading the required updates to the central aggregator for improving the global model. However, a key challenge is to maintain communication efficiency (i.e., the number of communications per iteration) when participating clients implement uncoordinated computation strategy during aggregation of model parameters. We formulate a utility maximization problem to tackle this difficulty, and propose a novel crowdsourcing framework, involving a number of participating clients with local training data to leverage FL. We show the incentivebased interaction between the crowdsourcing platform and the participating client's independent strategies for training a global learning model, where each side maximizes its own benefit. We formulate a two-stage Stackelberg game to analyze such scenario and find the game's equilibria. Further, we illustrate the efficacy of our proposed framework with simulation results. Results show that the proposed mechanism outperforms the heuristic approach with up to 22% gain in the offered reward to attain a level of target accuracy.
IEEE Internet of Things Journal, 2022
A recent take towards Federated Analytics (FA), which allows analytical insights of distributed d... more A recent take towards Federated Analytics (FA), which allows analytical insights of distributed datasets, reuses the Federated Learning (FL) infrastructure to evaluate the summary of model performances across the training devices. However, the current realization of FL adopts single servermultiple client architecture with limited scope for FA, which often results in learning models with poor generalization, i.e., an ability to handle new/unseen data, for real-world applications. Moreover, a hierarchical FL structure with distributed computing platforms demonstrates incoherent model performances at different aggregation levels. Therefore, we need to design a robust learning mechanism than the FL that (i) unleashes a viable infrastructure for FA and (ii) trains learning models with better generalization capability. In this work, we adopt the novel democratized learning (Dem-AI) principles and designs to meet these objectives. Firstly, we show the hierarchical learning structure of the proposed edge-assisted democratized learning mechanism, namely Edge-DemLearn, as a practical framework to empower generalization capability in support of FA. Secondly, we validate Edge-DemLearn as a flexible model training mechanism to build a distributed control and aggregation methodology in regions by leveraging the distributed computing infrastructure. The distributed edge computing servers construct regional models, minimize the communication loads, and ensure distributed data analytic application's scalability. To that end, we adhere to a nearoptimal two-sided many-to-one matching approach to handle the combinatorial constraints in Edge-DemLearn and solve it for fast knowledge acquisition with optimization of resource allocation and associations between multiple servers and devices. Extensive simulation results on real datasets demonstrate the effectiveness of the proposed methods. Index Terms-Federated analytics (FA), federated learning (FL), democratized learning (Dem-AI), multi-access edge computing (MEC).
IEEE Transactions on Wireless Communications, 2021
Federated Learning (FL) is a distributed learning framework that can deal with the distributed is... more Federated Learning (FL) is a distributed learning framework that can deal with the distributed issue in machine learning and still guarantee high learning performance. However, it is impractical that all users will sacrifice their resources to join the FL algorithm. This motivates us to study the incentive mechanism design for FL. In this paper, we consider a FL system that involves one base station (BS) and multiple mobile users. The mobile users use their own data to train the local machine learning model, and then send the trained models to the BS, which generates the initial model, collects local models and constructs the global model. Then, we formulate the incentive mechanism between the BS and mobile users as an auction game where the BS is an auctioneer and the mobile users are the sellers. In the proposed game, each mobile user submits its bids according to the minimal energy cost that the mobile users experiences in participating in FL. To decide winners in the auction and maximize social welfare, we propose the primal-dual greedy auction mechanism. The proposed mechanism can guarantee three economic properties, namely, truthfulness, individual rationality and efficiency. Finally, numerical results are shown to demonstrate the performance effectiveness of our proposed mechanism.
In this work, we design an elastic open radio access network (O-RAN) slicing for the industrial I... more In this work, we design an elastic open radio access network (O-RAN) slicing for the industrial Internet of things (IIoT). Unlike IoT, IIoT poses additional challenges such as severe communication environment, network-slice resource demand variations, and on-time information update from the IIoT devices during industrial production. First, we formulate the O-RAN slicing problem for on-time industrial monitoring and control where the objective is to minimize the cost of fresh information updates (i.e., age of information (AoI)) from the IIoT devices (i.e., sensors) while maintaining the energy consumption of those devices with the energy constraint as well as O-RAN slice isolation constraints. Second, we propose the intelligent ORAN framework based on game theory and machine learning to mitigate the problem’s complexity. We propose a two-sided distributed matching game in the O-RAN control layer that captures the IIoT channel characteristics and the IIoT service priorities to create ...
ArXiv, 2020
There is growing interest in applying distributed machine learning to edge computing, forming \em... more There is growing interest in applying distributed machine learning to edge computing, forming \emph{federated edge learning}. Compared with conventional distributed machine learning in a datacenter, federated edge learning faces non-independent and identically distributed (non-i.i.d.) and heterogeneous data, and the communications between edge workers, possibly through distant locations with unstable wireless networks, are more costly than their local computational overhead. In this work, we propose a distributed Newton-type algorithm (DONE) with fast convergence rate for communication-efficient federated edge learning. First, with strongly convex and smooth loss functions, we show that DONE can produce the Newton direction approximately in a distributed manner by using the classical Richardson iteration on each edge worker. Second, we prove that DONE has linear-quadratic convergence and analyze its computation and communication complexities. Finally, the experimental results with n...
2020 International Joint Conference on Neural Networks (IJCNN), 2020
ArXiv, 2020
This work proposes UE selection approaches to mitigate the straggler effect for federated learnin... more This work proposes UE selection approaches to mitigate the straggler effect for federated learning (FL) on cell-free massive multiple-input multiple-output networks. To show how these approaches work, we consider a general FL framework with UE sampling, and aim to minimize the FL training time in this framework. Here, training updates are (S1) broadcast to all the selected UEs from a central server, (S2) computed at the UEs sampled from the selected UE set, and (S3) sent back to the central server. The first approach mitigates the straggler effect in both Steps (S1) and (S3), while the second approach only Step (S3). Two optimization problems are then formulated to jointly optimize UE selection, transmit power and data rate. These mixed-integer mixed-timescale stochastic nonconvex problems capture the complex interactions among the training time, the straggler effect, and UE selection. By employing the online successive convex approximation approach, we develop a novel algorithm to ...
IEEE Transactions on Mobile Computing, 2021
Federated Learning is a new learning scheme for collaborative training a shared prediction model ... more Federated Learning is a new learning scheme for collaborative training a shared prediction model while keeping data locally on participating devices. In this paper, we study a new model of multiple federated learning services at the multi-access edge computing server. Accordingly, the sharing of CPU resources among learning services at each mobile device for the local training process and allocating communication resources among mobile devices for exchanging learning information must be considered. Furthermore, the convergence performance of different learning services depends on the hyper-learning rate parameter that needs to be precisely decided. Towards this end, we propose a joint resource optimization and hyper-learning rate control problem, namely MS À FEDL, regarding the energy consumption of mobile devices and overall learning time. We design a centralized algorithm based on the block coordinate descent method and a decentralized JP-miADMM algorithm for solving the MS À FEDL problem. Different from the centralized approach, the decentralized approach requires many iterations to obtain but it allows each learning service to independently manage the local resource and learning process without revealing the learning service information. Our simulation results demonstrate the convergence performance of our proposed algorithms and the superior performance of our proposed algorithms compared to the heuristic strategy.
IEEE Transactions on Wireless Communications, 2021
In this paper, we study the resource slicing problem in a dynamic multiplexing scenario of two di... more In this paper, we study the resource slicing problem in a dynamic multiplexing scenario of two distinct 5G services, namely Ultra-Reliable Low Latency Communications (URLLC) and enhanced Mobile BroadBand (eMBB). While eMBB services focus on high data rates, URLLC is very strict in terms of latency and reliability. In view of this, the resource slicing problem is formulated as an optimization problem that aims at maximizing the eMBB data rate subject to a URLLC reliability constraint, while considering the variance of the eMBB data rate to reduce the impact of immediately scheduled URLLC traffic on the eMBB reliability. To solve the formulated problem, an optimizationaided Deep Reinforcement Learning (DRL) based framework is proposed, including: 1) eMBB resource allocation phase, and 2) URLLC scheduling phase. In the first phase, the optimization problem is decomposed into three subproblems and then each subproblem is transformed into a convex form to obtain an approximate resource allocation solution. In the second phase, a DRL-based algorithm is proposed to intelligently distribute the incoming URLLC traffic among eMBB users. Simulation results show that our proposed approach can satisfy the stringent URLLC reliability while keeping the eMBB reliability higher than 90%.
IEEE Transactions on Communications, 2021
Ultra-reliable low-latency communication (uRLLC) and enhanced mobile broadband (eMBB) are two inf... more Ultra-reliable low-latency communication (uRLLC) and enhanced mobile broadband (eMBB) are two influential services of the emerging 5G cellular network. Latency and reliability are major concerns for uRLLC applications, whereas eMBB services claim for the maximum data rates. Owing to the tradeoff among latency, reliability and spectral efficiency, sharing of radio resources between eMBB and uRLLC services, heads to a challenging scheduling dilemma. In this paper, we study the co-scheduling problem of eMBB and uRLLC traffic based upon the puncturing technique. Precisely, we formulate an optimization problem aiming to maximize the minimum expected achieved rate (MEAR) of eMBB user equipment (UE) while fulfilling the provisions of the uRLLC traffic. We decompose the original problem into two sub-problems, namely scheduling problem of eMBB UEs and uRLLC UEs while prevailing objective unchanged. Radio resources are scheduled among the eMBB UEs on a time slot basis, whereas it is handled for uRLLC UEs on a mini-slot basis. Moreover, for resolving the scheduling issue of eMBB UEs, we use penalty successive upper bound minimization (PSUM) based algorithm, whereas the optimal transportation model (TM) is adopted for solving the same problem of uRLLC UEs. Furthermore, a heuristic algorithm is also provided to solve the Manuscript
IEEE Communications Letters, 2021
Currently, matching the incoming Internet of Things applications to the current state of computin... more Currently, matching the incoming Internet of Things applications to the current state of computing and networking resources of a mobile edge orchestrator (MEO) is critical for providing the high quality of service while temporally and spatially changing the incoming workload. However, MEO needs to scale its capacity concerning a large number of devices to avoid task failure and to reduce service time. To cope with this issue, we propose MEO with fuzzy-based logic that splits tasks from mobile devices and maps them onto the cloud and edge servers to reduce the latency of handling these tasks and task failures. A fuzzy-based MEO handles the multi-criteria decisionmaking process to decide where the offloaded task should run by considering multiple parameters in the same framework. Our approach selects the appropriate host for task execution and finds the optimal task-splitting strategy. Compared to the existing approaches, the service time using our proposal can achieve up to 7.6%, 22.6%, 38.9%, and 51.8% performance gains for augmented reality, healthcare, compute-intensive, and infotainment applications, respectively.
IEEE Transactions on Network and Service Management, 2021
In recent years, multi-access edge computing (MEC) is a key enabler for handling the massive expa... more In recent years, multi-access edge computing (MEC) is a key enabler for handling the massive expansion of Internet of Things (IoT) applications and services. However, energy consumption of a MEC network depends on volatile tasks that induces risk for energy demand estimations. As an energy supplier, a microgrid can facilitate seamless energy supply. However, the risk associated with energy supply is also increased due to unpredictable energy generation from renewable and non-renewable sources. Especially, the risk of energy shortfall is involved with uncertainties in both energy consumption and generation. In this paper, we study a risk-aware energy scheduling problem for a microgrid-powered MEC network. First, we formulate an optimization problem considering the conditional value-at-risk (CVaR) measurement for both energy consumption and generation, where the objective is to minimize the expected residual of scheduled energy for the MEC networks and we show this problem is an NP-hard problem. Second, we analyze our formulated problem using a multi-agent stochastic game that ensures the joint policy Nash equilibrium, and show the convergence of the proposed model. Third, we derive the solution by applying a multi-agent deep reinforcement learning (MADRL)-based asynchronous advantage actor-critic (A3C) algorithm with shared neural networks. This method mitigates the curse of dimensionality of the state space and chooses the best policy among the agents for the proposed problem. Finally, the experimental results establish a significant performance gain by considering CVaR for high accuracy energy scheduling of the proposed model than both the single and random agent models.
2019 IEEE Global Communications Conference (GLOBECOM), 2019
The nature of multi-access edge computing (MEC) is to deal with heterogeneous computational tasks... more The nature of multi-access edge computing (MEC) is to deal with heterogeneous computational tasks near to the end users, which induces the volatile energy consumption for the MEC network. As an energy supplier, a microgrid is able to enable seamless energy flow from renewable and non-renewable sources. In particular, the risk of energy demand and supply is increased due to nondeterministic nature of both energy consumption and generation. In this paper, we impose a risksensitive energy profiling problem for a microgrid-enabled MEC network, where we first formulate an optimization problem by considering Conditional Value-at-Risk (CVaR). Hence, the formulated problem can determine the risk of expected energy shortfall by coordinating with the uncertainties of both demand and supply, and we show this problem is NP-hard. Second, we design a multi-agent system that can determine a risk-sensitive energy profiling by coping with an optimal scheduling policy among the agents. Third, we devise the solution by applying a multi-agent deep reinforcement learning (MADRL) based on asynchronous advantage actor-critic (A3C) algorithm with shared neural networks. This approach mitigates the curse of dimensionality for state space and also, can admit the best energy profile policy among the agents. Finally, the experimental results establish the significant performance gain of the proposed model than that a single agent solution and achieves a high accuracy energy profiling with respect to risk constraint. Index Terms-green edge computing, microgrid, renewable energy, multi-agent deep reinforcement learning, conditional value-at-risk, energy profiling.
IEEE Transactions on Parallel and Distributed Systems
There is growing interest in applying distributed machine learning to edge computing, forming fed... more There is growing interest in applying distributed machine learning to edge computing, forming federated edge learning. Federated edge learning faces non-i.i.d. and heterogeneous data, and the communication between edge workers, possibly through distant locations and with unstable wireless networks, is more costly than their local computational overhead. In this work, we propose DONE, a distributed approximate Newton-type algorithm with fast convergence rate for communication-efficient federated edge learning. First, with strongly convex and smooth loss functions, DONE approximates the Newton direction in a distributed manner using the classical Richardson iteration on each edge worker. Second, we prove that DONE has linear-quadratic convergence and analyze its communication complexities. Finally, the experimental results with non-i.i.d. and heterogeneous data show that DONE attains a comparable performance to the Newton's method. Notably, DONE requires fewer communication iterations compared to distributed gradient descent and outperforms DANE and FEDL, state-of-the-art approaches, in the case of non-quadratic loss functions.
arXiv (Cornell University), Jun 15, 2020
Federated learning (FL) is a decentralized and privacy-preserving machine learning technique in w... more Federated learning (FL) is a decentralized and privacy-preserving machine learning technique in which a group of clients collaborate with a server to learn a global model without sharing clients' data. One challenge associated with FL is statistical diversity among clients, which restricts the global model from delivering good performance on each client's task. To address this, we propose an algorithm for personalized FL (pFedMe) using Moreau envelopes as clients' regularized loss functions, which help decouple personalized model optimization from the global model learning in a bi-level problem stylized for personalized FL. Theoretically, we show that pFedMe's convergence rate is state-of-the-art: achieving quadratic speedup for strongly convex and sublinear speedup of order 2/3 for smooth nonconvex objectives. Experimentally, we verify that pFedMe excels at empirical performance compared with the vanilla FedAvg and Per-FedAvg, a meta-learning based personalized FL algorithm.
IEEE Internet of Things Journal
Non-terrestrial networks (NTNs), which integrate space and aerial networks with terrestrial syste... more Non-terrestrial networks (NTNs), which integrate space and aerial networks with terrestrial systems, are a key area in the emerging sixth-generation (6G) wireless networks. As part of 6G, NTNs must provide pervasive connectivity to a wide range of devices, including smartphones, vehicles, sensors, robots, and maritime users. However, due to the high mobility and deployment of NTNs, managing the space-air-sea (SAS) NTN resources, i.e., energy, power, and channel allocation, is a major challenge. The design of a SAS-NTN for energy-efficient resource allocation is investigated in this study. The goal is to maximize system energy efficiency (EE) by collaboratively optimizing user equipment (UE) association, power control, and unmanned aerial vehicle (UAV) deployment. Given the limited payloads of UAVs, this work focuses on minimizing the total energy cost of UAVs (trajectory and transmission) while meeting EE requirements. A mixed-integer nonlinear programming problem is proposed, followed by the development of an algorithm to decompose, and solve each problem distributedly. The binary (UE association) and continuous (power, deployment) variables are separated using the Bender decomposition (BD), and then the Dinkelbach algorithm (DA) is used to convert fractional programming into an equivalent solvable form in the subproblem. A standard optimization solver is utilized to deal with the complexity of the master problem for binary variables. The alternating direction method of multipliers (ADMM) algorithm is used to solve the subproblem for the continuous variables. Our proposed algorithm provides a suboptimal solution, and simulation results demonstrate that the proposed algorithm achieves better EE than baselines.
arXiv (Cornell University), Apr 27, 2020
In recent years, the emerging topics of recommender systems that take advantage of natural langua... more In recent years, the emerging topics of recommender systems that take advantage of natural language processing techniques have attracted much attention, and one of their applications is the Conversational Recommender System (CRS). Unlike traditional recommender systems with contentbased and collaborative filtering approaches, CRS learns and models user's preferences through interactive dialogue conversations. In this work, we provide a summarization on the recent evolution of CRS, where deep learning approaches are applied to CRS and have produced fruitful results. We first analyze the research problems and present key challenges in the development of Deep Conversational Recommender Systems (DCRS), then present the current state of the field taken from the most recent researches, including the most common deep learning models that benefit DCRS. Finally, we discuss future directions of this vibrant area.
World Wide Web
In modern days, making recommendation for news articles poses a great challenge due to vast amoun... more In modern days, making recommendation for news articles poses a great challenge due to vast amount of online information. However, providing personalized recommendations from news articles, which are the sources of condense textual information is not a trivial task. A recommendation system needs to understand both the textual information of a news article, and the user contexts in terms of long-term and temporary preferences via the user’s historic records. Unfortunately, many existing methods do not possess the capability to meet such need. In this work, we propose a neural deep news recommendation model called CupMar, that not only is able to learn the user-profile representation in different contexts, but also is able to leverage the multi-aspects properties of a news article to provide accurate, personalized news recommendations to users. The main components of our CupMar approach include the News Encoder and the User-Profile Encoder. Specifically, the News Encoder uses multiple...
IEEE Transactions on Neural Networks and Learning Systems
Emerging cross-device artificial intelligence (AI) applications require a transition from convent... more Emerging cross-device artificial intelligence (AI) applications require a transition from conventional centralized learning systems toward large-scale distributed AI systems that can collaboratively perform complex learning tasks. In this regard, democratized learning (Dem-AI) lays out a holistic philosophy with underlying principles for building large-scale distributed and democratized machine learning systems. The outlined principles are meant to study a generalization in distributed learning systems that go beyond existing mechanisms such as federated learning (FL). Moreover, such learning systems rely on hierarchical self-organization of well-connected distributed learning agents who have limited and highly personalized data and can evolve and regulate themselves based on the underlying duality of specialized and generalized processes. Inspired by Dem-AI philosophy, a novel distributed learning approach is proposed in this article. The approach consists of a self-organizing hierarchical structuring mechanism based on agglomerative clustering, hierarchical generalization, and corresponding learning mechanism. Subsequently, hierarchical generalized learning problems in recursive forms are formulated and shown to be approximately solved using the solutions of distributed personalized learning problems and hierarchical Manuscript
IEEE Transactions on Communications, 2021
Unmanned aerial vehicles (UAVs) can provide an effective solution for improving the coverage, cap... more Unmanned aerial vehicles (UAVs) can provide an effective solution for improving the coverage, capacity, and the overall performance of terrestrial wireless cellular networks. In particular, UAV-assisted cellular networks can meet the stringent performance requirements of the fifth generation new radio (5G NR) applications. In this paper, the problem of energy-efficient resource allocation in UAV-assisted cellular networks is studied under the reliability and latency constraints of 5G NR applications. The framework of ruin theory is employed to allow solar-powered UAVs to capture the dynamics of harvested and consumed energies. First, the surplus power of every UAV is modeled, and then it is used to compute the probability of ruin of the UAVs. The probability of ruin denotes the vulnerability of draining out the power of a UAV. Next, the probability of ruin is used for efficient user association with each UAV. Then, power allocation for 5G NR applications is performed to maximize the achievable network rate using the water-filling approach. Simulation results demonstrate that the proposed ruin-based scheme can enhance the flight duration up to 61% and the number of served users in a UAV flight by up to 58%, compared to a baseline SINR-based scheme.
2019 IEEE Global Communications Conference (GLOBECOM), 2019
Federated learning (FL) rests on the notion of training a global model in a decentralized manner.... more Federated learning (FL) rests on the notion of training a global model in a decentralized manner. Under this setting, mobile devices perform computations on their local data before uploading the required updates to the central aggregator for improving the global model. However, a key challenge is to maintain communication efficiency (i.e., the number of communications per iteration) when participating clients implement uncoordinated computation strategy during aggregation of model parameters. We formulate a utility maximization problem to tackle this difficulty, and propose a novel crowdsourcing framework, involving a number of participating clients with local training data to leverage FL. We show the incentivebased interaction between the crowdsourcing platform and the participating client's independent strategies for training a global learning model, where each side maximizes its own benefit. We formulate a two-stage Stackelberg game to analyze such scenario and find the game's equilibria. Further, we illustrate the efficacy of our proposed framework with simulation results. Results show that the proposed mechanism outperforms the heuristic approach with up to 22% gain in the offered reward to attain a level of target accuracy.
IEEE Internet of Things Journal, 2022
A recent take towards Federated Analytics (FA), which allows analytical insights of distributed d... more A recent take towards Federated Analytics (FA), which allows analytical insights of distributed datasets, reuses the Federated Learning (FL) infrastructure to evaluate the summary of model performances across the training devices. However, the current realization of FL adopts single servermultiple client architecture with limited scope for FA, which often results in learning models with poor generalization, i.e., an ability to handle new/unseen data, for real-world applications. Moreover, a hierarchical FL structure with distributed computing platforms demonstrates incoherent model performances at different aggregation levels. Therefore, we need to design a robust learning mechanism than the FL that (i) unleashes a viable infrastructure for FA and (ii) trains learning models with better generalization capability. In this work, we adopt the novel democratized learning (Dem-AI) principles and designs to meet these objectives. Firstly, we show the hierarchical learning structure of the proposed edge-assisted democratized learning mechanism, namely Edge-DemLearn, as a practical framework to empower generalization capability in support of FA. Secondly, we validate Edge-DemLearn as a flexible model training mechanism to build a distributed control and aggregation methodology in regions by leveraging the distributed computing infrastructure. The distributed edge computing servers construct regional models, minimize the communication loads, and ensure distributed data analytic application's scalability. To that end, we adhere to a nearoptimal two-sided many-to-one matching approach to handle the combinatorial constraints in Edge-DemLearn and solve it for fast knowledge acquisition with optimization of resource allocation and associations between multiple servers and devices. Extensive simulation results on real datasets demonstrate the effectiveness of the proposed methods. Index Terms-Federated analytics (FA), federated learning (FL), democratized learning (Dem-AI), multi-access edge computing (MEC).
IEEE Transactions on Wireless Communications, 2021
Federated Learning (FL) is a distributed learning framework that can deal with the distributed is... more Federated Learning (FL) is a distributed learning framework that can deal with the distributed issue in machine learning and still guarantee high learning performance. However, it is impractical that all users will sacrifice their resources to join the FL algorithm. This motivates us to study the incentive mechanism design for FL. In this paper, we consider a FL system that involves one base station (BS) and multiple mobile users. The mobile users use their own data to train the local machine learning model, and then send the trained models to the BS, which generates the initial model, collects local models and constructs the global model. Then, we formulate the incentive mechanism between the BS and mobile users as an auction game where the BS is an auctioneer and the mobile users are the sellers. In the proposed game, each mobile user submits its bids according to the minimal energy cost that the mobile users experiences in participating in FL. To decide winners in the auction and maximize social welfare, we propose the primal-dual greedy auction mechanism. The proposed mechanism can guarantee three economic properties, namely, truthfulness, individual rationality and efficiency. Finally, numerical results are shown to demonstrate the performance effectiveness of our proposed mechanism.
In this work, we design an elastic open radio access network (O-RAN) slicing for the industrial I... more In this work, we design an elastic open radio access network (O-RAN) slicing for the industrial Internet of things (IIoT). Unlike IoT, IIoT poses additional challenges such as severe communication environment, network-slice resource demand variations, and on-time information update from the IIoT devices during industrial production. First, we formulate the O-RAN slicing problem for on-time industrial monitoring and control where the objective is to minimize the cost of fresh information updates (i.e., age of information (AoI)) from the IIoT devices (i.e., sensors) while maintaining the energy consumption of those devices with the energy constraint as well as O-RAN slice isolation constraints. Second, we propose the intelligent ORAN framework based on game theory and machine learning to mitigate the problem’s complexity. We propose a two-sided distributed matching game in the O-RAN control layer that captures the IIoT channel characteristics and the IIoT service priorities to create ...
ArXiv, 2020
There is growing interest in applying distributed machine learning to edge computing, forming \em... more There is growing interest in applying distributed machine learning to edge computing, forming \emph{federated edge learning}. Compared with conventional distributed machine learning in a datacenter, federated edge learning faces non-independent and identically distributed (non-i.i.d.) and heterogeneous data, and the communications between edge workers, possibly through distant locations with unstable wireless networks, are more costly than their local computational overhead. In this work, we propose a distributed Newton-type algorithm (DONE) with fast convergence rate for communication-efficient federated edge learning. First, with strongly convex and smooth loss functions, we show that DONE can produce the Newton direction approximately in a distributed manner by using the classical Richardson iteration on each edge worker. Second, we prove that DONE has linear-quadratic convergence and analyze its computation and communication complexities. Finally, the experimental results with n...
2020 International Joint Conference on Neural Networks (IJCNN), 2020
ArXiv, 2020
This work proposes UE selection approaches to mitigate the straggler effect for federated learnin... more This work proposes UE selection approaches to mitigate the straggler effect for federated learning (FL) on cell-free massive multiple-input multiple-output networks. To show how these approaches work, we consider a general FL framework with UE sampling, and aim to minimize the FL training time in this framework. Here, training updates are (S1) broadcast to all the selected UEs from a central server, (S2) computed at the UEs sampled from the selected UE set, and (S3) sent back to the central server. The first approach mitigates the straggler effect in both Steps (S1) and (S3), while the second approach only Step (S3). Two optimization problems are then formulated to jointly optimize UE selection, transmit power and data rate. These mixed-integer mixed-timescale stochastic nonconvex problems capture the complex interactions among the training time, the straggler effect, and UE selection. By employing the online successive convex approximation approach, we develop a novel algorithm to ...
IEEE Transactions on Mobile Computing, 2021
Federated Learning is a new learning scheme for collaborative training a shared prediction model ... more Federated Learning is a new learning scheme for collaborative training a shared prediction model while keeping data locally on participating devices. In this paper, we study a new model of multiple federated learning services at the multi-access edge computing server. Accordingly, the sharing of CPU resources among learning services at each mobile device for the local training process and allocating communication resources among mobile devices for exchanging learning information must be considered. Furthermore, the convergence performance of different learning services depends on the hyper-learning rate parameter that needs to be precisely decided. Towards this end, we propose a joint resource optimization and hyper-learning rate control problem, namely MS À FEDL, regarding the energy consumption of mobile devices and overall learning time. We design a centralized algorithm based on the block coordinate descent method and a decentralized JP-miADMM algorithm for solving the MS À FEDL problem. Different from the centralized approach, the decentralized approach requires many iterations to obtain but it allows each learning service to independently manage the local resource and learning process without revealing the learning service information. Our simulation results demonstrate the convergence performance of our proposed algorithms and the superior performance of our proposed algorithms compared to the heuristic strategy.
IEEE Transactions on Wireless Communications, 2021
In this paper, we study the resource slicing problem in a dynamic multiplexing scenario of two di... more In this paper, we study the resource slicing problem in a dynamic multiplexing scenario of two distinct 5G services, namely Ultra-Reliable Low Latency Communications (URLLC) and enhanced Mobile BroadBand (eMBB). While eMBB services focus on high data rates, URLLC is very strict in terms of latency and reliability. In view of this, the resource slicing problem is formulated as an optimization problem that aims at maximizing the eMBB data rate subject to a URLLC reliability constraint, while considering the variance of the eMBB data rate to reduce the impact of immediately scheduled URLLC traffic on the eMBB reliability. To solve the formulated problem, an optimizationaided Deep Reinforcement Learning (DRL) based framework is proposed, including: 1) eMBB resource allocation phase, and 2) URLLC scheduling phase. In the first phase, the optimization problem is decomposed into three subproblems and then each subproblem is transformed into a convex form to obtain an approximate resource allocation solution. In the second phase, a DRL-based algorithm is proposed to intelligently distribute the incoming URLLC traffic among eMBB users. Simulation results show that our proposed approach can satisfy the stringent URLLC reliability while keeping the eMBB reliability higher than 90%.
IEEE Transactions on Communications, 2021
Ultra-reliable low-latency communication (uRLLC) and enhanced mobile broadband (eMBB) are two inf... more Ultra-reliable low-latency communication (uRLLC) and enhanced mobile broadband (eMBB) are two influential services of the emerging 5G cellular network. Latency and reliability are major concerns for uRLLC applications, whereas eMBB services claim for the maximum data rates. Owing to the tradeoff among latency, reliability and spectral efficiency, sharing of radio resources between eMBB and uRLLC services, heads to a challenging scheduling dilemma. In this paper, we study the co-scheduling problem of eMBB and uRLLC traffic based upon the puncturing technique. Precisely, we formulate an optimization problem aiming to maximize the minimum expected achieved rate (MEAR) of eMBB user equipment (UE) while fulfilling the provisions of the uRLLC traffic. We decompose the original problem into two sub-problems, namely scheduling problem of eMBB UEs and uRLLC UEs while prevailing objective unchanged. Radio resources are scheduled among the eMBB UEs on a time slot basis, whereas it is handled for uRLLC UEs on a mini-slot basis. Moreover, for resolving the scheduling issue of eMBB UEs, we use penalty successive upper bound minimization (PSUM) based algorithm, whereas the optimal transportation model (TM) is adopted for solving the same problem of uRLLC UEs. Furthermore, a heuristic algorithm is also provided to solve the Manuscript
IEEE Communications Letters, 2021
Currently, matching the incoming Internet of Things applications to the current state of computin... more Currently, matching the incoming Internet of Things applications to the current state of computing and networking resources of a mobile edge orchestrator (MEO) is critical for providing the high quality of service while temporally and spatially changing the incoming workload. However, MEO needs to scale its capacity concerning a large number of devices to avoid task failure and to reduce service time. To cope with this issue, we propose MEO with fuzzy-based logic that splits tasks from mobile devices and maps them onto the cloud and edge servers to reduce the latency of handling these tasks and task failures. A fuzzy-based MEO handles the multi-criteria decisionmaking process to decide where the offloaded task should run by considering multiple parameters in the same framework. Our approach selects the appropriate host for task execution and finds the optimal task-splitting strategy. Compared to the existing approaches, the service time using our proposal can achieve up to 7.6%, 22.6%, 38.9%, and 51.8% performance gains for augmented reality, healthcare, compute-intensive, and infotainment applications, respectively.
IEEE Transactions on Network and Service Management, 2021
In recent years, multi-access edge computing (MEC) is a key enabler for handling the massive expa... more In recent years, multi-access edge computing (MEC) is a key enabler for handling the massive expansion of Internet of Things (IoT) applications and services. However, energy consumption of a MEC network depends on volatile tasks that induces risk for energy demand estimations. As an energy supplier, a microgrid can facilitate seamless energy supply. However, the risk associated with energy supply is also increased due to unpredictable energy generation from renewable and non-renewable sources. Especially, the risk of energy shortfall is involved with uncertainties in both energy consumption and generation. In this paper, we study a risk-aware energy scheduling problem for a microgrid-powered MEC network. First, we formulate an optimization problem considering the conditional value-at-risk (CVaR) measurement for both energy consumption and generation, where the objective is to minimize the expected residual of scheduled energy for the MEC networks and we show this problem is an NP-hard problem. Second, we analyze our formulated problem using a multi-agent stochastic game that ensures the joint policy Nash equilibrium, and show the convergence of the proposed model. Third, we derive the solution by applying a multi-agent deep reinforcement learning (MADRL)-based asynchronous advantage actor-critic (A3C) algorithm with shared neural networks. This method mitigates the curse of dimensionality of the state space and chooses the best policy among the agents for the proposed problem. Finally, the experimental results establish a significant performance gain by considering CVaR for high accuracy energy scheduling of the proposed model than both the single and random agent models.
2019 IEEE Global Communications Conference (GLOBECOM), 2019
The nature of multi-access edge computing (MEC) is to deal with heterogeneous computational tasks... more The nature of multi-access edge computing (MEC) is to deal with heterogeneous computational tasks near to the end users, which induces the volatile energy consumption for the MEC network. As an energy supplier, a microgrid is able to enable seamless energy flow from renewable and non-renewable sources. In particular, the risk of energy demand and supply is increased due to nondeterministic nature of both energy consumption and generation. In this paper, we impose a risksensitive energy profiling problem for a microgrid-enabled MEC network, where we first formulate an optimization problem by considering Conditional Value-at-Risk (CVaR). Hence, the formulated problem can determine the risk of expected energy shortfall by coordinating with the uncertainties of both demand and supply, and we show this problem is NP-hard. Second, we design a multi-agent system that can determine a risk-sensitive energy profiling by coping with an optimal scheduling policy among the agents. Third, we devise the solution by applying a multi-agent deep reinforcement learning (MADRL) based on asynchronous advantage actor-critic (A3C) algorithm with shared neural networks. This approach mitigates the curse of dimensionality for state space and also, can admit the best energy profile policy among the agents. Finally, the experimental results establish the significant performance gain of the proposed model than that a single agent solution and achieves a high accuracy energy profiling with respect to risk constraint. Index Terms-green edge computing, microgrid, renewable energy, multi-agent deep reinforcement learning, conditional value-at-risk, energy profiling.