A deep learning based soft interference cancellation symbol detector, based on the paper:

N. Shlezinger, R. Fu, and Y. C. Eldar.

“DeepSIC: Deep Soft Interference Cancellation for Multiuser MIMO Detection”.

arXiv preprint, arXiv:2002.03214, 2020.

The implementation of DeepSIC consists of two functions:

• GetDeepSICNet - generate and train DeepSIC MIMO detector. Training is carried out in a sequential manner (see Sequential Training in the above reference).

• s_fDetDeepSIC - use trained model to detect symbols, returns BER.

A code example for evaluating ViterbiNet can be found in the script DeepSIC_Test1.m.

This code requires Matlab with deep learning toolbox.