This paper investigates the optimal placement of a millimeter-wave (mmWave) base station (BS) within a realistic U-shaped environment with non-convex topology. The problem is challenging and NP-hard due to the non-convex topology and the non-convex objective functions which are the sum-rate maximization and max-min fairness, the latter being additionally non-smooth. To address this challenge, the BS placement is formulated as a Markov Decision Process (MDP). Then, we propose two deep reinforcement learning (DRL) techniques: First, the deployment area is discretized into a grid and optimized using a Deep Q-Network (DQN). Second, the U-shaped region is partitioned into continuous subspaces, where a Deep Deterministic Policy Gradient (DDPG) agent is dedicated to each subspace then the best BS placement is selected among partitions. Results demonstrate that optimal placement achieves full coverage and yields a Jain index of 0.99. Furthermore, the proposed partitioned multi-space DDPG achieves better solution than DQN with lower complexity.
Neuromorphic computing has so far been driven predominantly by machine-learning workloads, yet its underlying properties also make it particularly well suited to combinatorial optimization problems expressed in Ising or QUBO form. While neuromorphic Ising solvers have been demons…
Recent advances in neural rendering have unlocked unprecedented capabilities in 3D reconstruction and novel view synthesis, giving rise to applications such as virtual fly-throughs of a 3D scene reconstructed from a set of sparse, casually captured images. However, these renderin…
Compressed sensing accelerates MRI by reconstructing images from undersampled k-space, but performance depends strongly on sampling distribution. We propose an adaptive framework that selects Cartesian phase-encode lines sequentially using a fixed-cardinality quadratic unconstrai…
Mathematical optimization plays a fundamental role in signal processing and wireless communications, serving as an essential framework for the systematic design of modern systems. Many design challenges in these fields, as well as in many others, can naturally be formulated as op…
We consider the recovery of a pair of sparse vectors from a limited number of nonlinear observations of their superposition: $y_i=g(\inner{\ba_i}{\bPhi\bw^\ast+\bPsi\bz^\ast})+e_i$, $i=1,\dots,m$, with $m\ll n$, incoherent orthonormal bases $\bPhi,\bPsi$, a scalar link $g$, and n…
Graph neural networks (GNNs) have emerged as a promising approach to learning wireless policies efficiently by leveraging topology prior and incorporating relational inductive biases. However, when the optimal policy is not permutation equivariant (PE), conventional GNNs suffer f…