ClearNet
C + Learn + Net
A framework for the creation and training of vanilla and convolutional neural nets only depending on a C compiler and standard library.
Check out some of the examples in the examples directory.
Some Features
- Creation and training of vanilla and convolutional models
- Autodifferentiation engine
- Stochastic gradient descent and other optimization methods
- Saving and loading a model to a file
All of these functions are used in files in the examples directory.
- xor: Vanilla net creation and training on xor
- lin reg: Vanilla net creation and training on simple linear regression example
- iris: Vanilla net creation and training with stochastic gradient descent on the iris dataset
- full adder: Vanilla net creation and training with momentum on full adder operation
- mnist vanilla: Vanilla net creation and training with momentum and stochastic gradient descent on mnist dataset
- mnist mix: Creation and training of a convolutional net with dense, convolutional and pooling layers using momentum and stochastic gradient descent on the mnist dataset
- mnist convolutional: Creation and training of a convolutional net with convolutional and pooling layers using momentum and stochastic gradient descent on the mnist dataset