In this study, we address the challenge of packet based information routing in large-scale wireless communication networks. We approach this scenario by framing the problem as a statistical learning problem, where each node in the network relies only on the local data. Our exploration focuses on the idea of opportunistic routing, which exploits the broadcast nature of wireless communication to select the optimal relay node and transmit information packets to the destination node via multiple relay nodes. We present a distributed optimization method based on state augmentation (SA) that aims to maximize the total information on different source nodes of the network. Our formulation of the problem deploys graph neural networks (GNNs) to perform graph convolution on the topological connections between the network nodes. Using unsupervised learning, we derive optimal routing policies for the source nodes across multiple flows from the GNN output. Numerical results show the superiority of our proposed method by comparing a GNN model trained against standard algorithms.

While self-attention has been instrumental in the success of Transformers, it can lead to over-concentration on a few tokens during training, resulting in suboptimal information flow. Enforcing doubly-stochastic constraints in attention matrices has been shown to improve structure and balance in attention distributions. However, existing methods rely on iterative Sinkhorn normalization, which is computationally costly. In this paper, we introduce a novel, fully parallelizable doubly-stochastic attention mechanism based on sliced optimal transport, leveraging Expected Sliced Transport Plans (ESP). Unlike prior approaches, our method enforces doubly stochasticity without iterative Sinkhorn normalization, significantly enhancing efficiency. To ensure differentiability, we incorporate a temperature-based soft sorting technique, enabling seamless integration into deep learning models. Experiments across multiple benchmark datasets, including image classification, point cloud classification, sentiment analysis, and neural machine translation, demonstrate that our enhanced attention regularization consistently improves performance across diverse applications. Our implementation code can be found at \url{https://github.com/dariansal/ESPFormer}.

Network slicing is a key feature in 5G/NG cellular networks that creates customized slices for different service types with various quality-of-service (QoS) requirements, which can achieve service differentiation and guarantee service-level agreement (SLA) for each service type. In Wi-Fi networks, there is limited prior work on slicing, and a potential solution is based on a multi-tenant architecture on a single access point (AP) that dedicates different channels to different slices. In this paper, we define a flexible, constrained learning framework to enable slicing in Wi-Fi networks subject to QoS requirements. We specifically propose an unsupervised learning-based network slicing method that leverages a state-augmented primal-dual algorithm, where a neural network policy is trained offline to optimize a Lagrangian function and the dual variable dynamics are updated online in the execution phase. We show that state augmentation is crucial for generating slicing decisions that meet the ergodic QoS requirements.

We consider the problem of routing network packets in a large-scale communication system where the nodes have access to only local information. We formulate this problem as a constrained learning problem, which can be solved using a distributed optimization algorithm. We approach this distributed optimization using a novel state-augmentation (SA) strategy to maximize the aggregate information packets at different source nodes, leveraging dual variables corresponding to flow constraint violations. The construction is based on graph neural networks (GNNs) that employ graph convolutions over the underlying communication network topology. We devise an unsupervised learning algorithm to transform the output of the GNN architecture into optimal routing decisions. The proposed method takes advantage of only the local information available at each node and efficiently routes the desired packets to the destination. We provide numerical results demonstrating the superiority of the proposed method over baseline routing algorithms.

We consider the problem of power control in wireless networks, consisting of multiple transmitter-receiver pairs communicating with each other over a single shared wireless medium. To achieve both a high total rate and a level of fairness across users, we formulate a policy optimization problem with constraints on the minimum per-user rate across network configurations with an adaptive slack parameter. To apply unsupervised learning algorithms in the dual domain, we parameterize the power control policy, slack variable, and dual parameters using graph neural networks (GNNs), which leverage the network topology to create a scalable and network-invariant processing architecture. We use a primal- dual algorithm to learn the optimal GNN parameters and demonstrate via numerical simulations the resulting GNNs' success in achieving the right balance between sum- and 5thpercentile rates throughout a range of network configurations.

We consider the pool-based active learning problem, where only a subset of the training data is labeled, and the goal is to query a batch of unlabeled samples to be labeled so as to maximally improve model performance. We formulate the problem using constrained learning, where a set of constraints bounds the performance of the model on labeled samples. Considering a primal-dual approach, we optimize the primal variables, corresponding to the model parameters, as well as the dual variables, corresponding to the constraints. As each dual variable indicates how significantly the perturbation of the respective constraint affects the optimal value of the objective function, we use it as a proxy of the informativeness of the corresponding training sample. Our approach, which we refer to as Active Learning via Lagrangian dualitY, or ALLY, leverages this fact to select a diverse set of unlabeled samples with the highest estimated dual variables as our query set. We demonstrate the benefits of our approach in a variety of classification and regression tasks and discuss its limitations depending on the capacity of the model used and the degree of redundancy in the dataset. We also examine the impact of the distribution shift induced by active sampling and show that ALLY can be used in a generative mode to create novel, maximally-informative samples.

We propose a new approach for power control in wireless networks using self-supervised learning. We partition a multi-layer perceptron that takes as input the channel matrix and outputs the power control decisions into a backbone and a head, and we show how we can use contrastive learning to pre-train the backbone so that it produces similar embeddings at its output for similar channel matrices and vice versa, where similarity is defined in an information-theoretic sense by identifying the interference links that can be optimally treated as noise. The backbone and the head are then fine-tuned using a limited number of labeled samples. Simulation results show the effectiveness of the proposed approach, demonstrating significant gains over pure supervised learning methods in both sum-throughput and sample efficiency.1

Learning representations from sets has become increasingly important with many applications in point cloud processing, graph learning, image/video recognition, and object detection. We introduce a geometrically-interpretable and generic pooling mechanism for aggregating a set of features into a fixed-dimensional representation. In particular, we treat elements of a set as samples from a probability distribution and propose an end-to-end trainable Euclidean embedding for sliced-Wasserstein distance to learn from set-structured data effectively. We evaluate our proposed pooling method on a wide variety of set-structured data, including point-cloud, graph, and image classification tasks, and demonstrate that our proposed method provides superior performance over existing set representation learning approaches. Our code is available at https://github.com/navid-naderi/PSWE.

We propose a mechanism for distributed resource management and interference mitigation in wireless networks using multi-agent deep reinforcement learning (RL). We equip each transmitter in the network with a deep RL agent that receives delayed observations from its associated users, while also exchanging observations with its neighboring agents, and decides on which user to serve and what transmit power to use at each scheduling interval. Our proposed framework enables agents to make decisions simultaneously and in a distributed manner, unaware of the concurrent decisions of other agents. Moreover, our design of the agents' observation and action spaces is scalable, in the sense that an agent trained on a scenario with a specific number of transmitters and users can be applied to scenarios with different numbers of transmitters and/or users. Simulation results demonstrate the superiority of our proposed approach compared to decentralized baselines in terms of the tradeoff between average and 5thpercentile user rates, while achieving performance close to, and even in certain cases outperforming, that of a centralized information-theoretic baseline. We also show that our trained agents are robust and maintain their performance gains when experiencing mismatches between train and test deployments.

For millimeter wave (mmWave) systems with large-scale arrays, hybrid processing structure is usually used at both transmitters and receivers to reduce the complexity and cost, which poses a very challenging issue in channel estimation, especially at the low transmit signal-to-noise ratio regime. In this paper, deep convolutional neural network (CNN) is employed to perform wideband channel estimation for mmWave massive multiple-input multiple-output (MIMO) systems. In addition to exploiting spatial correlation, our joint channel estimation approach also exploits the frequency correlation, where the tentatively estimated channel matrices at multiple adjacent subcarriers are input into the CNN simultaneously. The complexity analysis and numerical results show that the proposed CNN based joint channel estimation outperforms the non-ideal minimum mean-squared error (MMSE) estimator with reduced complexity and achieves the performance close to the ideal MMSE estimator. It is also quite robust to different propagation scenarios.

A typical wireless transceiver includes a radio frequency integrated circuit (RFIC) and a baseband modem (BBIC) which are connected through an input/output (I/O) interface. The wide-bandwidth and high-rate millimeter wave (mmWave) systems put a heavy burden on the power dissipation of the I/O interface of a transceiver. In this paper, a novel low power fully digital architecture with blind beam tracking and spatial compression (FDA-BTSC) is introduced to reduce the rate and power dissipation of the I/O interface. Spatial compression of the received signal is feasible due to the sparsity of mmWave channels. An efficient spatial compression is realized through codebook-based beamforming and fast blind beam tracking. Provided analysis and evaluations show that the proposed architecture is potentially as power efficient as existing analog and hybrid mmWave architectures. In addition, FDA-BTSC significantly drops the baseband processing complexity and power consumption level to the same order as hybrid beamforming, while it maintains the advantages of the fully digital beamforming in terms of low latency of the beam management and high efficiency of the digital beamforming.

We consider the problem of interference management in wireless cellular networks with caches at both base stations and receivers, and we characterize the degrees of freedom (DoFs) per cell to within an additive gap of (1/3) and a multiplicative gap of 2 for all system parameters, under one-shot linear schemes. Our result indicates that the one-shot linear DoF per cell scales linearly with the total amount of cache available in the cell, i.e., the sum of the caches at the central base station and all the receivers within the cell, resembling a similar phenomenon previously observed for the case of fully connected wireless networks. To establish the result, we propose a decentralized and randomized cache placement and a delivery scheme, which, on one hand, utilizes the overlap of contents at the base station caches to zero-force part of their outgoing interference and, on the other hand, uses the receiver cache contents to create coded multicasting opportunities, so that the receivers can eliminate the remaining interference due to undesired packets. We also provide a converse argument, which shows that our achievable one-shot linear DoF per cell is within a constant additive gap of (1/3) and a multiplicative gap of 2 of its optimum.

We present a method for parallel dynamic cell selection and link scheduling in order to simultaneously activate a subset of non- interfering transmitter-receiver links to manage the interference in an ultra-dense network setting. We show how we can utilize a practical feedback method for periodic reporting of channel state information measured at the users back to the network to enable implementing our interference management scheme. Through simulation results, we demonstrate the superiority of our proposed interference management scheme over several benchmark schemes in terms of sum-rate and network coverage.

We consider the problem of interference management in wireless cellular networks with caches at both base stations and receivers and we characterize the degrees-of-freedom (DoF) per cell to within an additive gap of 1 and a multiplicative gap of 2 for all system parameters, under one-shot linear schemes. Our result indicates that the one-shot linear DoF per cell scales linearly with the total amount of cache that is available within each cell. A similar phenomenon had been previously observed for the case of fully-connected wireless networks. Hence, our result demonstrates that it also holds in cellular networks, despite the presence of path loss and fading which results in partial connectivity of the network topology. To establish the result, we propose a randomized and decentralized cache placement and a delivery scheme which, on one hand, utilizes the overlap of contents at base station caches to zero-force part of their outgoing interference, and on the other hand, uses the receiver caches to create coded multicasting opportunities, so that the receivers can eliminate the remaining interference due to undesired packets. We also provide a converse argument to show that our achievable one-shot linear DoF per cell is within an additive gap of 1 and a multiplicative gap of 2 of its optimal value.

We consider a system, containing a library of multiple files and a general memoryless communication network through which a server is connected to multiple users, each equipped with a local isolated cache of certain size that can be used to store part of the library. Each user will ask for one of the files in the library, which needs to be delivered by the server through the intermediate communication network. The objective is to design the cache placement (without prior knowledge of users' future requests) and the delivery phase in order to minimize the (normalized) delivery delay. We assume that the delivery phase consists of two steps: (1) generation of a set of multicast messages at the server, one for each subset of users, and (2) delivery of the multicast messages to the users. In this setting, we show that there exists a universal scheme for cache placement and multicast message generation, which is independent of the underlying communication network between the server and the users, and achieves the optimal delivery delay to within a constant factor for all memoryless networks. We prove this result, even though the capacity region of the underlying communication network is not known, even approximately. This result shows that in the aforementioned setting, a separation between caching and multicast message generation on one hand, and delivering the multicast messages to the users on the other hand is approximately optimal. This result has the important practical implication that the prefetching can be done independent of network structure in the upcoming delivery phase.

We propose a method for interference mitigation in an ultra-dense wireless network scenario in 5G, where a group of transmit points (TPs) intend to serve multiple user equipments (UEs) using the same wireless resource. The proposed UE scheduling subroutine may schedule a single UE or multiple UEs for each TP, where multiple UEs can be served via non-orthogonal multiple-access (NOMA). Afterwards, the link scheduling subroutine utilizes an extended form of the optimality conditions for treating interference as noise in single-user networks for the case where NOMA is also involved. This subroutine determines an activation/deactivation pattern of TPs while maintaining fairness among competing users and minimizing the level of interference among concurrent transmissions. We show that our proposed interference management scheme can improve over the benchmark schemes both in terms of sum-throughput and coverage, striking the right trade-off between these two performance metrics.

We study the symmetric degrees-of-freedom (DoF) of partially connected interference networks under linear coding strategies without channel state information at the transmitters beyond topology. We assume that the receivers are equipped with reconfigurable antennas that can switch among their preset modes. In such a network setting, we characterize the class of network topologies in which half linear symmetric DoF is achievable. Moreover, we derive two general upper bounds on the linear symmetric DoF for arbitrary network topologies. We also demonstrate the tightness of our bounds for the class of network topologies in which all the transmitters in the network have at most two co-interferers.

We consider the problem of non-coherent interference alignment, in which the goal is to align the signals of multiple interfering transmitters at a single receiver where the transmitters are not aware of the channel state information. We cast this problem as a problem of determining rank loss conditions for a column concatenation of full-rank matrices, such that each row of the composing matrices is scaled by a random coefficient. We determine necessary and sufficient conditions for the design of each matrix, such that the random ensemble will almost surely lose rank by a certain amount. The result is proved by converting the problem to determining rank loss conditions for the union of some specific matroids, and then using tools from matroid and graph theories to derive the necessary and sufficient conditions. As an application, we discuss how this result can be applied to the problem of topological interference management, and characterize the linear symmetric degrees of freedom for a class of network topologies.

We consider a system, comprising a library of N files (e.g., movies) and a wireless network with a KTtransmitters, each equipped with a local cache of size of MTfiles and a KRreceivers, each equipped with a local cache of size of MRfiles. Each receiver will ask for one of the N files in the library, which needs to be delivered. The objective is to design the cache placement (without prior knowledge of receivers' future requests) and the communication scheme to maximize the throughput of the delivery. In this setting, we show that the sum degrees-of-freedom (sum-DoF) of {(KTMT+KRMR)/N, KR} is achievable, and this is within a factor of 2 of the optimum, under uncoded prefetching and one-shot linear delivery schemes. This result shows that (i) the one-shot sum-DoF scales linearly with the aggregate cache size in the network (i.e., the cumulative memory available at all nodes), (ii) the transmitters' caches and receivers' caches contribute equally in the one-shot sum-DoF, and (iii) caching can offer a throughput gain that scales linearly with the size of the network. To prove the result, we propose an achievable scheme that exploits the redundancy of the content at transmitter's caches to cooperatively zero-force some outgoing interference, and availability of the unintended content at the receiver's caches to cancel (subtract) some of the incoming interference. We develop a particular pattern for cache placement that maximizes the overall gains of cache-aided transmit and receive interference cancellations. For the converse, we present an integer optimization problem which minimizes the number of communication blocks needed to deliver any set of requested files to the receivers. We then provide a lower bound on the value of this optimization problem, hence leading to an upper bound on the linear one-shot sum-DoF of the network, which is within a factor of 2 of the achievable sum-DoF.

We consider a system, comprising a library of files (e.g., movies) and a wireless network with an arbitrary number of transmitters and receivers, where each node is equipped with a local cache memory. The system operates in two phases, the prefetching phase, where each cache is pre-populated from the contents of the library, up to its limited size, and then the delivery phase, where each receiver reveals its request for a file from the library, and the system needs to deliver the requested files. The objective is to design the cache placement and the communication scheme to maximize the rate of delivery for arbitrary set of requested files. We characterize the sum degrees-of-freedom (sum-DoF) of this network to within a factor of 2 for all system parameters, under one-shot linear schemes. In particular, we show that the linear sum-DoF scales linearly with the aggregate cache size in the network (i.e., the cumulative memory available at all nodes). The proposed achievable scheme exploits the redundancy of the content at transmitters' caches to cooperatively zero-force some outgoing interference, and availability of the unintended content at the receivers' caches to cancel (subtract) some of the incoming interference. The outer bound is derived by an optimization argument which bounds the number of communication blocks needed to deliver any requested contents to the receivers. This result demonstrates that in this setting, caches at the transmitters' side are equally valuable as the caches at the receivers' side. In addition, it shows that caching can offer a throughput gain that scales linearly with the size of the network.

We study the symmetric degrees-of-freedom (DoF) of partially connected interference networks under linear coding strategies without channel state information at the transmitters beyond topology. We assume that the receivers are equipped with reconfigurable antennas that can switch among their preset modes. In such a network setting, we characterize the class of network topologies in which half linear symmetric DoF is achievable. Moreover, we derive a general upper bound on the linear symmetric DoF for arbitrary network topologies. We also show that this upper bound is tight if the transmitters have at most two co-interferers.

We study the problem of interference management in fast fading wireless networks, in which the transmitters are only aware of network topology. We consider a class of retransmission-based schemes, where transmitters in the network are only allowed to resend their symbols in order to assist with the neutralization of interference at the receivers. We introduce a necessary and sufficient condition on the network topology, under which half symmetric degrees-of-freedom (DoF) is achievable through the considered retransmission-based schemes. This corresponds to the “best” topologies since half symmetric DoF is the highest possible value for the symmetric DoF in the presence of interference. We show that when the condition is satisfied, there always exists a set of carefully chosen transmitters in the network, such that by retransmission of their symbols at an appropriate time slot, we can neutralize all the interfering signals at the receivers. Quite surprisingly, we also show that for any given network topology, if we cannot achieve half symmetric DoF by retransmission-based schemes, then there does not exist any linear scheme that can do so. We also consider a practical network scenario that models cell edge users in a heterogeneous network, and show that the characterized condition on the network topology occurs frequently. Furthermore, we numerically evaluate the achievable rates of the DoF-optimal retransmission-based scheme in such network scenario, and show that its throughput gain is not restricted to the asymptotic DoF analysis.

We consider the problem of determining rank loss conditions for a concatenation of full-rank matrices, such that each row of the composing matrices is scaled by a random coefficient. This problem has applications in wireless interference management and recommendation systems. We determine necessary and sufficient conditions for the design of each matrix, such that the random ensemble will almost surely lose rank by a certain amount. The result is proved by converting the problem to determining rank loss conditions for the union of some specific matroids, and then using tools from matroid and graph theories to derive the necessary and sufficient conditions. As an application, we discuss how this result can be applied to the problem of topological interference management, and characterize the linear symmetric degrees of freedom for a class of network topologies.

It is shown that in the K-user interference channel, if for each user the desired signal strength is no less than the sum of the strengths of the strongest interference from this user and the strongest interference to this user (all values in decibel scale), then the simple scheme of using point-to-point Gaussian codebooks with appropriate power levels at each transmitter and treating interference as noise (TIN) at every receiver (in short, TIN scheme) achieves all points in the capacity region to within a constant gap. The generalized degrees of freedom (GDoF) region under this condition is a polyhedron, which is shown to be fully achieved by the same scheme, without the need for time-sharing. The results are proved by first deriving a polyhedral relaxation of the GDoF region achieved by TIN, and then providing a dual characterization of this polyhedral region via the use of potential functions, and finally proving the optimality of this region in the desired regime.

We consider partially connected K -user interference networks, where the transmitters have no knowledge about the channel gain values, but they are aware of network topology. We introduce several linear algebraic and graph theoretic concepts to derive new topology-based outer bounds and inner bounds on the symmetric degrees-of-freedom of these networks. We evaluate our bounds for two classes of networks to demonstrate their tightness for most networks in these classes, quantify the gain of our inner bounds over benchmark interference management strategies, and illustrate the effect of network topology on these gains.

We consider the problem of spectrum sharing in device-to-device communication systems. Based on the recently-found condition for the optimality of treating interference as noise, we introduce the new notion of information-theoretic independent set (ITIS) which denotes a subset of users in a wireless network that are information-theoretically eligible to transmit data at the same time. This leads to our novel spectrum sharing scheme called information-theoretic link scheduling (ITLinQ) which at each time slot schedules the users in a single ITIS to transmit simultaneously. We perform a capacity analysis of ITLinQ and in a network model with a random placement of nodes in a cell, we characterize a lower bound on the fraction the information theoretic capacity region that it is able to achieve to within a gap. We will show that ITLinQ can outperform the conventional independent set scheduling by a multiplicative gain that scales with the number of users. We also present a distributed way of implementing ITLinQ and show that it yields a sum-rate gain of higher than 100% over similar state-of-the-art spectrum sharing mechanisms, such as FlashLinQ.

We consider the problem of spectrum sharing in device-to-device communication systems. Inspired by the recent optimality condition for treating interference as noise, we define a new concept of information-theoretic independent sets (ITISs), which indicates the sets of links for which simultaneous communication and treating the interference from each other as noise is information-theoretically optimal (to within a constant gap). Based on this concept, we develop a new spectrum sharing mechanism, called information-theoretic link scheduling (ITLinQ), which at each time schedules those links that form an ITIS. We first provide a performance guarantee for ITLinQ by characterizing the fraction of the capacity region that it can achieve in a network with sources and destinations randomly located within a fixed area. Furthermore, we demonstrate how ITLinQ can be implemented in a distributed manner, using an initial two-phase signaling mechanism that provides the required channel state information at all the links. Through numerical analysis, we show that distributed ITLinQ can outperform similar state-of-the-art spectrum sharing mechanisms, such as FlashLinQ, by more than 100% of sum-rate gain, while keeping the complexity at the same level. Finally, we discuss a variation of the distributed ITLinQ scheme, which can also guarantee fairness among the links in the network and numerically evaluate its performance.

We study the symmetric degrees-of-freedom (DoF) of partially connected K-user interference networks in which the transmitters are unaware of the actual channel gain values. Several linear algebraic and graph theoretic concepts are introduced to derive new outer and inner bounds on the symmetric DoF for arbitrary network topologies. We evaluate the bounds for a class of networks, showing that the bounds are tight for most topologies in that class, and we also quantify the gains obtained over benchmark schemes.