Learning Optimal Resource Allocations in Wireless Systems

Eisen, Mark; Zhang, Clark; Chamon, Luiz F. O.; Lee, Daniel D.; Ribeiro, Alejandro

doi:10.1109/tsp.2019.2908906

Cited by 219 publications

(199 citation statements)

References 42 publications

Supporting

Mentioning

198

Contrasting

Order By: Relevance

“…There nonetheless remains the practical challenge of training models that can meet the scale of modern wireless systems. Fully connected neural networks (FCNNs) may seem appealing due to their well known universal approximation property [11], [18]. However, FCNNs are also well known to be unworkable except in small scale problems.…”

Section: Introductionmentioning

confidence: 99%

Optimal Wireless Resource Allocation With Random Edge Graph Neural Networks

Eisen

Ribeiro

2020

IEEE Trans. Signal Process.

Self Cite

215

191

View full text Add to dashboard Cite

We consider the problem of optimally allocating resources across a set of transmitters and receivers in a wireless network. The resulting optimization problem takes the form of constrained statistical learning, in which solutions can be found in a model-free manner by parameterizing the resource allocation policy. Convolutional neural networks architectures are an attractive option for parameterization, as their dimensionality is small and does not scale with network size. We introduce the random edge graph neural network (REGNN), which performs convolutions over random graphs formed by the fading interference patterns in the wireless network. The REGNN-based allocation policies are shown to retain an important permutation equivariance property that makes them amenable to transference to different networks. We further present an unsupervised model-free primal-dual learning algorithm to train the weights of the REGNN. Through numerical simulations, we demonstrate the strong performance REGNNs obtain relative to heuristic benchmarks and their transference capabilities.

show abstract

Section: Introductionmentioning

confidence: 99%

Optimal Wireless Resource Allocation With Random Edge Graph Neural Networks

Eisen

Ribeiro

2020

IEEE Trans. Signal Process.

Self Cite

215

191

View full text Add to dashboard Cite

show abstract

“…To find the solution of constrained functional optimization problems, unsupervised learning techniques are proposed in [6,8,9]. A constrained optimization problem can be transformed into an unconstrained problem by using the Lagrangian approach.…”

Section: B Functional Optimization Using Unsupervised Learning and Rmentioning

confidence: 99%

“…This saddle point can be found by optimizing the policy f (h) and the Lagrangian multipliers for maximizing and minimizing the Langrangian function, respectively. We note that in contrast to the functional optimization considered in [8], P2 also takes general instantaneous constraints into consideration whose associated multipliers in P3 of Fig. 1 are actually functions of the environment's status [6].…”

Section: B Learning Generic Functional Optimization Without Labelsmentioning

confidence: 99%

“…Given the observations of the OF's and CFs' value, the finite difference method can be used for estimating their gradients [8], which is however inefficient when these functions have high dimensions and hence they are not applicable to functional optimization problems having infinite dimensions.…”

Section: A Learning Deterministic Policymentioning

confidence: 99%

“…Stochastic policies are widely used in RL for trading off exploration against exploitation [11]. For functional optimization, the policy network can also parameterize a stochastic policy, which is applicable to both discrete and continuous policy learning [8]. To realize a stochastic policy, the policy network is designed to take the environment's status as its input and then output the probability distribution of the action to be executed.…”

Section: B Learning Stochastic Policymentioning

confidence: 99%

See 2 more Smart Citations

Optimizing Wireless Systems Using Unsupervised and Reinforced-Unsupervised Deep Learning

et al. 2020

View full text Add to dashboard Cite

Resource allocation and transceivers in wireless networks are usually designed by solving optimization problems subject to specific constraints, which can be formulated as variable or functional optimization. If the objective and constraint functions of a variable optimization problem can be derived, standard numerical algorithms can be applied for finding the optimal solution, which however incur high computational cost when the dimension of the variable is high. To reduce the online computational complexity, learning the optimal solution as a function of the environment's status by deep neural networks (DNNs) is an effective approach. DNNs can be trained under the supervision of optimal solutions, which however, is not applicable to the scenarios without models or for functional optimization where the optimal solutions are hard to obtain. If the objective and constraint functions are unavailable, reinforcement learning can be applied to find the solution of a functional optimization problem, which is however not tailored to optimization problems in wireless networks. In this article, we introduce unsupervised and reinforced-unsupervised learning frameworks for solving both variable and functional optimization problems without the supervision of the optimal solutions. When the mathematical model of the environment is completely known and the distribution of environment's status is known or unknown, we can invoke unsupervised learning algorithm. When the mathematical model of the environment is incomplete, we introduce reinforcedunsupervised learning algorithms that learn the model by interacting with the environment. Our simulation results confirm the applicability of these learning frameworks by taking a user association problem as an example.

show abstract

AI‐Enabled Radio Access Network

Jang

Lee

2019

Wiley 5G Ref

View full text Add to dashboard Cite

Initialization and performance optimization of wireless networks are key enablers for the fifth‐generation (5G) radio access network (RAN) design. However, heterogeneity of wireless devices and massive connectivity requirements on 5G systems result in intractable end‐to‐end communication models where traditional model‐based wireless protocols cannot be straightforwardly applied. To tackle this issue, a new 5G RAN design paradigm is addressed based on artificial intelligence (AI) techniques, which lead to data‐driven optimization solutions for nontrivial implementation problems of 5G systems. Still, however, it has not been adequately investigated how to include AI technologies into 5G RAN architecture. To answer this fundamental question, this article illustrates the possibility and the feasibility of using AI for handling implementation difficulties occurring in 5G communication systems. The way to replace traditional model‐based communication solutions with AIs is investigated, especially focused on massive random access and nonconvex resource management applications. Several examples of reinforcement learning, supervised learning, and unsupervised learning techniques applied to 5G RAN are visited. Recent successes of AI‐defined 5G RAN verify that AIs can effectively identify unknown and optimal communication rules in various networking scenarios including initial accessing protocols for massive connectivity and wireless resource control strategies for network performance optimization.

show abstract

Learning Optimal Resource Allocations in Wireless Systems

Cited by 219 publications

References 42 publications

Optimal Wireless Resource Allocation With Random Edge Graph Neural Networks

Optimal Wireless Resource Allocation With Random Edge Graph Neural Networks

Optimizing Wireless Systems Using Unsupervised and Reinforced-Unsupervised Deep Learning

AI‐Enabled Radio Access Network

Contact Info

Product

Resources

About