Title: Enhance belief propagation decoding using ensemble learning approaches

Authors: Md Aleem; Sharjeel Ahmad; Syed Jalal Ahmad

Addresses: Department of ECE, KG Reddy College of Engineering and Technology, Hyderabad, 501504, India ' Department of CSE, Malla Reddy Engineering College, Hyderabad, 500100, India ' Department of ECE, Kashmir College of Engineering and Technology (KCET), Jammu & Kashmir, 190001, India

Abstract: Combining multiple learning models is an effective strategy for improving performance in complex decision-making systems. In belief propagation (BP) decoding, methods such as belief propagation list (BPL) and weighted belief propagation (WBP) provide complementary advantages in accuracy and adaptability. This work presents a novel decoding framework, termed ensemble weighted belief propagation (E-WBP), which integrates ensemble learning principles into the BP paradigm. The proposed method constructs multiple WBP decoders trained with diverse weight configurations during an offline phase. These decoders are then combined into a unified architecture that enhances decoding robustness. Performance results show that, for a fixed number of iterations, the proposed approach outperforms conventional BP decoding and approaches neural network-assisted WBP schemes, achieving an approximate gain of 0.3 dB over BPL at a frame error rate of 2 × 10^-3. Furthermore, offline training reduces computational complexity and storage overhead, enabling efficient deployment in practical communication systems.

Keywords: integrated learning; individual learner; polar code; BP decoding; weighted BP decoding.

DOI: 10.1504/IJNVO.2025.153494

International Journal of Networking and Virtual Organisations, 2025 Vol.33 No.4, pp.342 - 360

Received: 19 Jun 2024
Accepted: 13 Jun 2025
Published online: 11 May 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article