An Efficient Algorithm for Capacity-Approaching Noisy Adaptive Group Testing

Jonathan Scarlett

DOI: https://doi.org/10.48550/arXiv.1911.02764

2019-11-07

Abstract:In this paper, we consider the group testing problem with adaptive test designs and noisy outcomes. We propose a computationally efficient four-stage procedure with components including random binning, identification of bins containing defective items, 1-sparse recovery via channel codes, and a "clean-up" step to correct any errors from the earlier stages. We prove that the asymptotic required number of tests comes very close to the best known information-theoretic achievability bound (which is based on computationally intractable decoding), and approaches a capacity-based converse bound in the low-sparsity regime.

Computation,Information Theory,Probability

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to efficiently conduct group testing through adaptive test design in the presence of noise, in order to identify a small number of "defective" items in a large set. Specifically, the author proposes a four - stage process, including random binning, identifying bins containing defective items, achieving 1 - sparse recovery through channel coding, and a "clean - up" step to correct possible errors in the previous stages. This method is computationally efficient, and the number of tests it requires is close to the known information - theoretic achievability bounds, especially close to the converse bound of the capacity baseline in the low - sparsity case.

### Core Problems of the Paper
1. **Group Testing in a Noisy Environment**: How to efficiently identify a small number of defective items when the test results are affected by noise?
2. **Adaptive Test Design**: How to design tests in an adaptive manner so that the test results can be reliably used to recover the set of defective items?
3. **Algorithm Efficiency and Performance Guarantee**: How to design a computationally efficient algorithm whose performance almost matches the known best information - theoretic achievability bounds?

### Solutions
The author proposes a four - stage adaptive group - testing algorithm:
1. **Random Binning**: Randomly divide all items into multiple bins.
2. **Identify Bins Containing Defective Items**: Through non - adaptive group - testing methods, identify which bins contain defective items.
3. **1 - Sparse Recovery**: Within the identified bins, use standard channel - coding techniques for 1 - sparse recovery to determine individual defective items.
4. **Clean - up Step**: Correct possible errors in the previous stages to ensure that the final estimation results are accurate.

### Main Contributions
- **Close to the Optimal Information - Theoretic Bounds**: The proposed algorithm is very close to the known best information - theoretic achievability bounds in terms of the number of tests required.
- **Computational Efficiency**: The algorithm is computationally efficient and suitable for practical applications.
- **Theoretical Analysis**: Provides a detailed theoretical analysis to prove the effectiveness and performance guarantee of the algorithm.

### Formula Summary
- **Log - Likelihood Ratio**: \[ n \geq \frac{k \log \frac{p}{k}}{\log 2 - H_2(\rho)} (1 - o(1)) \]
where \( H_2(\rho)=\rho \log \frac{1}{\rho}+(1 - \rho) \log \frac{1}{1 - \rho} \).
- **Number of Tests**: \[ n=\left( \frac{k \log \frac{p}{k}}{\log 2 - H_2(\rho)}+\frac{k \log k}{D(\rho \| 1 - \rho)} \right) (1 + o(1)) \]

Through these methods and formulas, the author has successfully solved the problem of efficiently conducting group testing in a noisy environment.