Simple library for creating sequential models of convolutional neural networks. Library is for the solving small problems because is aimed on the readable code and modularity more than performance.
Install-Package Convsharp -Version 1.0.0
dotnet add package Convsharp --version 1.0.0
<PackageReference Include="Convsharp" Version="1.0.0" />
paket add Convsharp --version 1.0.0
First release of library:
- 1D and 2D convolutional layer
- 1D and 2D maxpooling layer
- flatten layer
- linear layer
- activation layer (ReLU, Tanh, Sigmoid, Softmax)
- loss functions (binary cross-entropy, categorical cross-entropy, MSE)
- optimalization algorithms (mini-batch SGD, Adam)
- regularization techniques: Dropout, L2 regularization
- Veldrid.ImageSharp (>= 4.1.4)
This package is not used by any popular GitHub repositories.