Concrete MAP Detection: A Machine Learning Inspired Relaxation

Authors: E. Beck, C. Bockelmann, A. Dekorsy
Abstract:

Motivated by large linear inverse problems where the complexity of the Maximum A-Posteriori (MAP) detector grows exponentially with system dimensions, e.g., large MIMO, we introduce a method to relax a discrete MAP problem into a continuous one. The relaxation is inspired by recent ML research and offers many favorable properties reflecting its quality. Hereby, we derive an iterative detection algorithm based on gradient descent optimization: Concrete MAP Detection (CMD). We show numerical results of application in large MIMO systems that demonstrate superior performance w.r.t. all considered State of the Art approaches.

Document type: Conference Paper
Publication: Hamburg, Germany, 18. - 20. February 2020
Conference: 24th International ITG Workshop on Smart Antennas (WSA 2020)
Volume: 24
Files:
Concrete MAP Detection: A Machine Learning Inspired Relaxation
WSA20_ConcreteMAPDetection_EdgarBeck_final.pdf316 KB
BibTEX
Last change on 29.01.2020 by E. Beck
AIT ieee tzi ith Fachbereich 1
© Department of Communications Engineering - University of BremenImprint / Contact