Sparse Foundations: Achieving Capacity with Polylogarithmic LDGM Polar Codes

Capacity-achieving Polar-based LDGM Codes with Crowdsourcing Applications

2020-06-01
James Chin-Jen Pang, Hessam Mahdavifar, S. Sandeep Pradhan
Summary
Problem
Method
Results
Takeaways
Abstract

The paper introduces a novel construction of capacity-achieving codes based on polar codes with strictly constrained generator matrix column weights. By utilizing a "splitting algorithm" and large polarization kernels, the authors achieve capacity over Binary-input Memoryless Symmetric (BMS) channels with column weights upper bounded by , significantly improving encoding sparsity.

TL;DR

Encoding complexity and task assignment overhead are critical bottlenecks in distributed systems and crowdsourcing. This paper presents a breakthrough in Low-Density Generator Matrix (LDGM) codes by proving that we can achieve Shannon capacity over any binary-input memoryless symmetric (BMS) channel while keeping the weight of every column in the generator matrix restricted to a polylogarithmic function of the block length: .

The Sparsity Paradox in Coding Theory

In the world of error correction, we frequently encounter LDPC (Low-Density Parity-Check) codes, which are famous for their sparse parity-check matrices. However, their dual counterparts, LDGM codes, have been historical underdogs.

The problem? Many LDGM constructions suffer from high error floors or fail to be "asymptotically good." Yet, in applications like crowdsourcing, LDGM sparsity is non-negotiable. If you are assigning tasks to human workers (XOR queries on items), you cannot give one worker 1,000 items to compare; the "column weight" of your assignment matrix must be small.

Methodology: Polarization Meets Splitting

The authors utilize Polar Codes as their engine. Standard Arıkan polar codes (using the kernel) have a recursive Kronecker structure that leads to some columns having weights as high as or even in specific submatrices.

1. Large Kernels and Identity Expansion

Instead of the standard matrix, the authors consider kernels. They define a new generator structure: This creates a block-diagonal-like structure that facilitates capacity analysis while allowing for massive parallelization.

2. The Splitting Algorithm (The Secret Sauce)

The most innovative part of this work is how they handle "heavy" columns. If a column's Hamming weight exceeds a threshold , the algorithm "splits" it into columns such that their sum in equals the original column, but each individual new column is light.

需替换为架构图 Note: The splitting process effectively increases the number of columns (slightly reducing the rate) but maintains the ability to recover the original information bits via Successive Cancellation (SC) decoding.

Experimental Analysis: Thresholds and Rates

The paper rigorously proves that the rate loss incurred by splitting columns vanishes as grows, provided the threshold is set correctly.

  • The Critical Constant: They identify a constant .
  • Performance: For , the code remains capacity-achieving.
  • Most Common Weight: They show that in these transformed polar codes, the "Most Common Column Weight" naturally concentrates around logarithmic values.

实验结果对比 Table II in the paper shows that kernels like and actually provide better sparsity orders than the kernels optimized purely for error exponents.

Crowdsourcing Application: Dealing with Unreliable Workers

The authors apply their construction to a Binary Symmetric Channel (BSC) model of crowdsourcing. In this scenario, workers might provide incorrect labels with probability . By concatenating an LDPC code (for decompression) with their sparse LDGM-polar code, they create a query scheme where:

  1. The number of queries approaches the information-theoretic limit.
  2. Each worker is only assigned a very small, manageable number of items.

Critical Insight & Conclusion

This paper is a significant theoretical contribution. It validates the conjecture that column weights polynomially sublinear in are sufficient for capacity. More importantly, it provides an explicit construction for polylogarithmic sparsity.

Limitations: While theoretically sound, the "Splitting Algorithm" increases the block length. In practical, finite-length regimes, the constant factors hidden in the notation might still be significant for very noisy channels.

Future Work: The next frontier is extending this sparsity to non-symmetric channels and exploring the hardware implementation of the splitting-based SC decoder.

Find Similar Papers

Try Our Examples

  • Search for recent papers that extend the column-weight splitting algorithm to non-binary or higher-order Galois field polar codes.
  • What are the state-of-the-art results for the scaling exponent of polar codes using large kernels as mentioned in the Korada et al. (2010) foundations?
  • Find studies applying these sparse LDGM-based polar codes specifically to distributed machine learning gradients or secure multiparty computation.
Contents
Sparse Foundations: Achieving Capacity with Polylogarithmic LDGM Polar Codes
1. TL;DR
2. The Sparsity Paradox in Coding Theory
3. Methodology: Polarization Meets Splitting
3.1. 1. Large Kernels and Identity Expansion
3.2. 2. The Splitting Algorithm (The Secret Sauce)
4. Experimental Analysis: Thresholds and Rates
5. Crowdsourcing Application: Dealing with Unreliable Workers
6. Critical Insight & Conclusion