Work-In-Progress autograd deep learning library made in java.
As of 8/13/2021
- Implementation of
DenseLayer
- Implementation of
MSELoss
- Implementation of
Relu
- Implementation of
LeakyRelu
- Implementation of
Sigmoid
- Batch training loop
Need to add more activation functions, different kinds of optimizers, as well as classes that can update weight matricies independently of the model's gradient update function, many other layers(like cnn, lstm, etc), and more. (certainly will have to change my code quite a bit(as my gradient update function won't work for softmax and layers like CNN's and RNN's))
Also have to change my training loop a bit...
This readme will be updated with more information, such as installation/usage/etc.
main.java contains a simple example of this.
This is far from done, and certainly not ready for production usage. Try DL4J
Perhaps some code was not implemented correctly(this project is a way for me to learn the low-level aspects of deep learning). If that is the case, please report an issue explaining so.