# JSCC

**Repository Path**: zhoub86/JSCC

## Basic Information

- **Project Name**: JSCC
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-11-19
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# JSCC

Run **train.py** for BSC or AWGN channel

Run **train_OFDM.py** for multipath channel based on OFDM system

So far, can only work with **CIFAR-10** dataset. Currently working to include **CelebA**

Three GANs available:  **vanilla GAN, LSGAN, WGAN**

**BSC channel**:

* Soft and hard Gunbel Softmax relaxed Bernoulli distribution

**OFDM system**:

* LS, LMMSE channel estimation
* ZF, MMSE, implicit equalization
* A version of feedback of CSI included

## visdom compatible 

* ssh port forwarding from server to local machine `ssh -N -f -L localhost:8998:localhost:8998 username@server` 
* Then view the training dynamics in `localhost:8998` with local browser

Use **nohup** to run multiple threads

The basic coding framework is based on pix2pix [Github](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix)


## 08/16 update

* Support multiple pilots for channel estimation
* Added three forward methods
* Able to test the OFDM channel in models/test_OFDM.py

## 08/25 update

* Fixed a bug in OFDM system
* Modified the clipping layer
* Modified LLR calculation for baseline
* Added QPSK symbols as pilots
* Added scripts to train pure GAN 

## 08/29 update

* Add residual connections 

## 09/05 update

* Two kinds of pilots (QPSK, ZadoffChu)
* New LLR calculation for baseline clipping

## 09/28 Structures

![Networks](networks.png)