Neural Network Channel Coding for Small-Block Wireless Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing channel codes, such as Reed-Muller (RM) and Polar codes, are inefficient for small and medium block lengths in wireless communication, lacking computational efficiency and reliability, and there is a need for new codes that can improve average-case reliability beyond their worst-case pairwise distance focus.
Innovation Solution
A new family of non-linear codes, called KO codes, are developed using neural networks to generalize the Kronecker operation underlying algebraic codes, trained with a Plotkin tree structure to enhance encoding and decoding processes, leveraging both algebraic and pseudorandom constructions for improved reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If algebraic codes (RM, Polar) are used for small and medium block lengths, then computational efficiency is maintained, but reliability is insufficient compared to random codes
Solution Approach 1:
The code construction is segmented into modular components: Plotkin tree structure provides the hierarchical framework, while neural network layers are inserted at specific nodes to introduce non-linearity. This segmentation allows the system to maintain the computational efficiency of algebraic codes while incorporating the reliability benefits of non-linear transformations.
Solution Approach 2:
The patent creates a composite code structure by combining algebraic code elements (Plotkin tree, Kronecker products) with neural network components. This composite approach integrates the structured efficiency of algebraic codes with the adaptive power of machine learning models, achieving both computational efficiency and high reliability.
2Reliability
If random codes are used to achieve information theoretic optimality, then reliability is improved, but encoding and decoding become computationally inefficient
Solution Approach 1:
Instead of making the entire code random, the patent applies non-linear neural network transformations only at specific locations (nodes) within the Plotkin tree structure. This localized application of complexity maintains overall computational efficiency while introducing sufficient non-linearity to achieve reliability close to random codes.
Solution Approach 2:
The neural network layers act as intermediaries between the structured algebraic code framework and the desired random-code-like reliability. These intermediary components transform the deterministic algebraic operations into probabilistic-like behavior without fully abandoning the efficient algebraic structure.
3Reliability
If non-linear transformations are introduced to generalize Kronecker operation, then reliability beyond worst-case pairwise distance is achieved, but computational complexity increases
Solution Approach 1:
The patent applies non-linear neural network transformations partially, only where needed within the code structure, rather than throughout the entire encoding and decoding process. This partial application achieves the necessary average-case reliability improvement while minimizing the increase in computational complexity.
Data Source
AI summary
A method of encoding a set of information bits to produce a codeword that encodes the set of information bits for reliable communication is provided. The set of information bits is received. The set of information bits are provided to a plurality of permutation layers separated by neural network processing layers. Each permutation layer accepts an input vector and generates a reordered output vector that is a reordering of the input vector. Each neural network processing layer accepts a vector of input values and generates a vector of output values based on a non-linear function of the vector of input values. The reordered output vector of a final permutation layer of the plurality of permutation layers is provided as the codeword. In some embodiments, a corresponding method of decoding a codeword to retrieve a set of information bits is provided.