Gumbel-Softmax Discretization Constraint, Differentiable IDS Channel, and an IDS-Correcting Code for DNA Storage

2024-07-10

AI Summary

Efficient error correction for insertion, deletion, and substitution (IDS) errors remains a major challenge in DNA-based data storage. Method: This paper proposes THEA-code—the first end-to-end trainable, channel-customized IDS error-correcting code framework. It innovatively integrates Gumbel-Softmax relaxation for discrete code optimization with a differentiable IDS channel model, overcoming the non-differentiability and poor generalizability of conventional constructive coding schemes. Built upon an autoencoder architecture, THEA-code jointly optimizes codeword generation, channel simulation, and decoding. Contribution/Results: Experiments demonstrate that THEA-code significantly improves decoding accuracy and robustness across diverse, realistic IDS channel models. It supports flexible adaptation to varying channel characteristics without manual code redesign, establishing a novel, learnable paradigm for efficient error correction in DNA storage.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Security and Privacy: Large-scale security measurements

Abstract

Insertion, deletion, and substitution (IDS) error-correcting codes have garnered increased attention with recent advancements in DNA storage technology. However, a universal method for designing IDS-correcting codes across varying channel settings remains underexplored. We present an autoencoder-based method, THEA-code, aimed at efficiently generating IDS-correcting codes for complex IDS channels. In the work, a Gumbel-Softmax discretization constraint is proposed to discretize the features of the autoencoder, and a simulated differentiable IDS channel is developed as a differentiable alternative for IDS operations. These innovations facilitate the successful convergence of the autoencoder, resulting in channel-customized IDS-correcting codes with commendable performance across complex IDS channels.

Problem

• DNA storage
• error correction codes
• reliability enhancement

Innovation

• THEA-code
• Gumbel-Softmax optimization
• DNA information storage