Project 61: Machine Learning for Optical Communications

Contact Information:

Prof. Lilin Yi

Project Description and Objectives:

Machine learning and neural networks have become very popular these years and have shown their strength especially in the domain of computer vision and machine translation. The neural network also comes into view of optical communities with more layers and a more intrinsic inter-layer relationship. A much more powerful tool, convolutional neural network (CNN), is now widely used in the domain of computer vision and also is the key for AlphaGo to defeat various professional Go players. CNN has also shown its powerful capability in optical performance monitoring and modulation formats identification.

This project mainly focuses on how machine learning can solve the signal performance distortion in optical fiber transmission, including dispersion, nonlinearities, and bandwidth limitation-induced inter-symbol interference. The performance of different machine learning structures such as supported-vector machine (SVM), fully-connected neuron network, CNN, and recurrent neuron network (RNN) will be compared and evaluated.

Eligibility Requirements:

Interested students should have basic knowledge of optical communications and programming.

Main Tasks:

Finish a research report.

Give two research presentations (1. Background review, 2. Technical progress).

Submit a paper to a conference or a journal as a co-author.

Website:

Lab: http://front.sjtu.edu.cn/~llyi/index_en.html

School: www.seiee.sjtu.edu.cn/