Channel decoding via machine learning

Error-correcting codes (codes) are the backbone of the modern information age and were essential to the invention of groundbreaking technology such as WiFi, cellular, cable, and satellite modems. In this thesis, we focus on using machine learning techniques to design efficient and reliable data-driven decoders for state-of-the-art channel codes. In the first half of the thesis, we introduce a neural-augmented decoder for Turbo codes called TinyTurbo. TinyTurbo has complexity comparable to the classical max-log-MAP algorithm but has much better reliability than the max-log-MAP baseline and performs close to the MAP algorithm. We show that TinyTurbo exhibits strong robustness on a variety of practical channels of interest, such as EPA and EVA channels, which are included in the LTE standards. We also show that TinyTurbo strongly generalizes across different rate, blocklengths, and trellises. In the second half of the thesis, we focus on designing data-driven decoders for the polar code family: Polar codes and PAC codes. We pose the decoding of PAC codes as a tree-search problem, and introduce PAC- DQN, a reinforcement learning based decoder. While PAC-DQN achieves a near-optimal reliability for short codes, it suffers from poor training sample complexity and is not scalable to larger codes. We then introduce CRISP, a GRU-powered neural decoder, which uses curriculum learning to achieve excellent reliability and scale to larger codes. We show that CRISP out-performs the successive-cancellation (SC) decoder and attains near-optimal reliability performance on the Polar(16, 32), Polar(22, 64) and PAC(16, 32) codes.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Hebbar, Shibara Ashwin

Contributors dc:contributor

• Viswanath, Pramod

• communication
• machine learning
• deep learning
• reinforcement learning
• channel coding
• error correction coding

Statement dc:rights

• Copyright 2022 Shibara Ashwin Hebbar

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/116261

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F116261