Generate code for a neural network that can be trained on an FPGA

[LeoBuron]

As a application engineer I want to train my neural network during runtime on the FPGA.

In this user story the minimal steps for the training on FPGAs are presented:

General

• [ ] #355
• [ ] #356

Stub-Skeleton

• [ ] #357
• [ ] #358

Forward neural network

Linear

• [ ] #359

ReLU

• [ ] Check if relu implementation can be used directly

Gradient Calculation neural network

Linear

• [ ] Implement the gradient calculation of the linear layer
• [ ] Implement the backward propagation algorithm for the linear layer

ReLU

• [ ] Implement the backward propagation algorithm for the linear layer

Optimizer

• [ ] Implement a basic SGD without momentum to update the weights

Loss-function

• [ ] Implement the MSE loss function