Higher-order Interpretations of Deepcode, a Learned Feedback Code

TL;DR

This paper provides a clear analytical interpretation of Deepcode, a learned feedback error-correcting code, highlighting its higher-order error correction capabilities through a two-stage decoding process.

Contribution

It introduces a succinct analytical framework for Deepcode, revealing its higher-order error correction mechanism with learned parameters and feedback.

Findings

Demonstrates higher-order error correction in Deepcode

Provides analytical encoder and decoder expressions

Shows two-stage decoding process enhances error correction

Abstract

We present an interpretation of Deepcode, a learned feedback code that showcases higher-order error correction relative to an earlier interpretable model. By interpretation, we mean succinct analytical encoder and decoder expressions (albeit with learned parameters) in which the role of feedback in achieving error correction is easy to understand. By higher-order, we mean that longer sequences of large noise values are acted upon by the encoder (which has access to these through the feedback) and used in error correction at the decoder in a two-stage decoding process.