# ComplexValuedCNN

**Repository Path**: wang-changan/ComplexValuedCNN

## Basic Information

- **Project Name**: ComplexValuedCNN
- **Description**: Complex Valued CNN
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-12-08
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# ComplexValuedCNN

Complex CNN in Pytorch, implementation of the paper Deep Complex Networks : https://arxiv.org/abs/1705.09792


### Modules implemeted : 
- [x] Complex Conv2d ![conv](https://raw.githubusercontent.com/Medabid1/ComplexValuedCNN/master/img/nn.png?token=AHFNLWOQ7J6TJR6IXL7AQFS5MQO5M)
- [x] Complex ConvTranspose2d
- [x] Complex AvgPool2d
- [x] Complex Linear

### Activation function : 
- [x] zRelu 
- [x] CRelu

## Usage :
1. Define your Complex input using the class CPLX.
    `x = CPLX(real_part, imaginary_part)`
2. Build your model using src modules.
3. Use the loss function in `utils.py` to train your model. if the labels are real valued, set `use_magnitude = True`

#### ToDo
-  Complex UNet