Deep Unsupervised Learning for Optimization With Box and Monotone Matrix Based Polytope Constraints: A Case-Study of D2D Wireless Networks

Acharjee, Bindubritta; Hanif, Muhammad; Waqar, Omer

doi:10.1109/lwc.2023.3316114

Cited by 2 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Remark: Both meta-learning and GNNs can be applied for a variety of wireless problems to reduce the training complexity for adaptation, since meta-learning methods are agnostic to problems while GNNs can leverage the permutation properties that are widely existed in wireless policies. In addition, they are compatible with the techniques to satisfy non-convex constraints, say the methods proposed in [33] and [34]. To validate this, we consider the same problem as in [34], i.e., optimizing power control to maximize the sum rate for D2D networks under the transmit power constraint and SINR constraints.…”

Section: Discussionmentioning

confidence: 96%

“…In addition, they are compatible with the techniques to satisfy non-convex constraints, say the methods proposed in [33] and [34]. To validate this, we consider the same problem as in [34], i.e., optimizing power control to maximize the sum rate for D2D networks under the transmit power constraint and SINR constraints. We incorporate MAML and GNNs with the method in [34], which introduces two scaling operations on the outputs of the DNN by leveraging the property of monotone matrices, to facilitate quick adaptation meanwhile satisfying both constraints.…”

Section: Discussionmentioning

confidence: 96%

“…To validate this, we consider the same problem as in [34], i.e., optimizing power control to maximize the sum rate for D2D networks under the transmit power constraint and SINR constraints. We incorporate MAML and GNNs with the method in [34], which introduces two scaling operations on the outputs of the DNN by leveraging the property of monotone matrices, to facilitate quick adaptation meanwhile satisfying both constraints. Our simulation results show that This work is licensed under a Creative Commons Attribution 4.0 License.…”

Section: Discussionmentioning

confidence: 99%

“…The two ways of introducing inductive biases for dealing with the mismatch issue are compatible, and both are compatible with the techniques to address other issues (e.g., the methods proposed in [33] and [34] satisfy complicated constraints). Yet a majority of previous works using metalearning for wireless problems employed FNNs or CNNs, disregarding the PE properties of wireless policies.…”

Section: Introductionmentioning

confidence: 97%

See 3 more Smart Citations

Meta-Learning for Wireless Communications: A Survey and a Comparison to GNNs

Zhao,

Wu,

et al. 2024

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

Deep learning has been used for optimizing a multitude of wireless problems. Yet most existing works assume that training and test samples are drawn from the same distribution, which is not true in dynamic environments. As a result, a deep neural network (DNN) may require retraining with newly collected samples to adapt to a new scenario. The retraining complexity can be reduced by introducing inductive biases, which can either be learned automatically or be embedded in DNNs manually. For example, inductive biases can be learned from related tasks by meta-learning in the form of good initializations, reusable modules and feature extractors, or be introduced by designing structural DNNs such as graph neural networks (GNNs) with appropriate permutation equivariance (PE) properties. All previous works on meta-learning overlooked the prior-known PE properties, which widely exist in wireless policies and can be leveraged to improve sample efficiency. This article reviews meta-learning for wireless communications and compares it with GNNs trained for a single task. We first introduce meta-learning by comparing it with conventional learning, analyze its inductive biases, and review its applications for radio transmission. We then compare meta-learning with GNNs trained for a single task by taking beamforming problem as an example to show their pros and cons in solving the mismatch issues. Simulation results show that GNNs are sample efficient while meta-learning can reduce the time complexity for adaptation. Both the considered GNNs and meta-learning methods are inefficient for adapting to environments with time-varying problem scales.

show abstract

Section: Discussionmentioning

confidence: 96%