Related Experiment Video
Updated: May 13, 2026
07:12
Using Informational Connectivity to Measure the Synchronous Emergence of fMRI Multi-voxel Information Across Time
Published on: July 1, 2014
12.3K
Enhanced mutual information neural estimators for optical fiber communication
Optics Letters
|August 2, 2024
Summary
Researchers developed an enhanced Mutual Information Neural Estimator (E-MINE) for optical communications. This deep learning approach accurately estimates mutual information in nonlinear fiber channels, improving system performance.
Area of Science:
- Optical Communications
- Information Theory
- Machine Learning
Background:
- Accurate mutual information (MI) estimation is crucial for optical communication systems.
- Existing methods struggle with nonlinear optical fiber channels due to unknown models.
- Additive white Gaussian noise (AWGN) channel MI estimation is well-established but insufficient for complex channels.
Purpose of the Study:
- Introduce a novel Mutual Information Neural Estimator (MINE) for optical fiber communications.
- Propose an enhanced MINE (E-MINE) to improve estimation accuracy and stability.
- Evaluate E-MINE's performance in AWGN and nonlinear optical fiber channels.
Main Methods:
- Developed an enhanced Mutual Information Neural Estimator (E-MINE).
- Increased training batch size in E-MINE for improved accuracy and stability.
- Compared E-MINE with traditional Monte Carlo (MC) methods.
Main Results:
- E-MINE achieved highly accurate MI estimations in AWGN channels.
- E-MINE demonstrated strong consistency with symbol-by-symbol MI estimations in long-haul fiber channels.
- Multi-symbol estimation with E-MINE resulted in a 0.16 bits/4D-symbol improvement.
Conclusions:
- E-MINE offers a robust solution for MI estimation in challenging optical communication environments.
- The enhanced deep learning approach shows significant potential for optimizing communication system design.
- This work paves the way for advanced deep learning applications in optical communications.
