Open 9il opened 7 years ago
This is not priority. Matrix inversion is not ready yet
I have updated gemm
API in the code.
Also, I've separated mir.experimental.model.rbf
and mir.experimental.ml.rbfann
modules - rbf
would be nothrow @nogc
low-level API (I tried following mir.glas
style here), and rbfann
would contain artificial neural network model implementation that relies on rbf
module.
I cannot start working on Mir PR until we get matrix inversion, but @9il, if you'd like, take a quick look at current state, I could use a rough feedback.
Also, I'm keeping it single threaded not to involve std.parallelism
. I suppose we'll solve this eventually...