This is an educational project for Java users to learn basic neural networks. The structure of the library, which is mainly inspired by pytorch, is kept as minimal as possible, so that the learner can easily understand the structure as well as play with the library (e.g. adding layers, optimizers).
Note that:
-
The library is designed without optimization in mind. Speed is never a target.
-
It is strongly suggested that the learner starts with a basic material about neural networks. This tutorial is a great start.
-
Following pytorch and several other deep learning libraries, an instance is represented by a row vector. Therefore, a linear layer is Y = XW + b rather than Y = WX + b.
- jblas (tested with v. 1.2.5)
Please find a number of tutorials in the tutorial folder
A Javadoc can be found here