| PhD Seminar

Name of the Speaker: Mr. Ravi Kumar (EE20D004)
Guide: Dr. Manivasakan R
Venue: ESB-244 (Seminar Hall)
Date/Time: 5th May 2025 (Monday), 4 PM
Title: DPM-Solver-2M: A Fast Multistep DPM-Solver-Based Scheme for Real-Time MIMO Channel Estimation

Abstract :

Real-time multiple-input multiple-output (MIMO) channel estimation poses a significant challenge due to stringent low-latency and high-accuracy demands in emerging wireless communication systems. To address this, we propose a novel multistep fast Ordinary Differential Equation (ODE)-based diffusion probabilistic model (DPM) approach, named DPM-Solver-2M, specifically tailored for MIMO channels. Recent advances in diffusion probabilistic models (DPMs) have outperformed traditional generative models like VAEs and GANs, providing enhanced stability and superior sample generation capabilities. Unlike conventional discrete-time diffusion models, our framework reformulates the lightweight discrete noise prediction network into a continuous-time trajectory, enabling the application of fast high-order numerical solvers such as DPM-Solver++ . This reformulation facilitates substantial reduction in the number of inference steps while preserving estimation accuracy, achieving approximately a 4× speedup in terms of number of function evaluations (NFE) compared to Markovian or ancestral sampling-based DPM estimators. Building upon advancements from score-based generative modeling and efficient ODE sampling frameworks, our method maintains low computational complexity due to sparse angle-delay domain channel, while ensuring rapid convergence within 10–15 solver steps.