This is a pet probject playing with generative modelling and the VALA programming language.
Think of TensorFlow rewritten in VALA.
VALA does not support operator overriding. So we will focus on expression the operations as a computing chart. YAML is probably not the best language for this (e.g. feel pylearn2)
We will first implementing numpy functionalities in dask way; and most functionalities of autograd.
We will not look too much into distributed computing. The plan is we will probably make use of DBUS. That's far down the road.
This is an unfunded project.
We use meson to build the vala project.
On Fedora, the following commands shall work
mkdir build cd build meson .. ninja-build
- To declare an array inline, use new int[]{1, 2, 3, }.