Nikolay Zakirov
Nikolay Zakirov
After reading the source code, I decided that the problem is related to xinerama. Although not being able to solve this in a beautiful way, here is a script that...
Thanks a lot, derf. There is one more thing that I did not mention. When the script is run the screen is off. It is being switched off with ```...
I took a brief look at this. SAGA/SAG/SG solvers take advantage of the fact that the loss function (e.g. least-squares or cross entropy loss) is a **sum** of more simple...
Actually now I am thinking that if we add `df_i` method - how to compute the gradient of the i-th term to `BinaryObjectiveFunction`, `MultiClassObjectiveFunction` then the SAGA solver implementation can...
I would start with the vanila SAG/SAGA algorithm. ``` // epoch loop // loop through samples // extract a random sample // update weights // get current prediction // compute...
Any objections to this crate then - https://docs.rs/sprs/latest/sprs/ ? It is CSR. The algo would work this way then. The optimizer would accept normal Matrices and Vectors from this lib....
Ok, great. Working on it
Anyone solved this issue yet or maybe found a workaround?
I don't know if this might help at all. I tried to replace the OPTIONAL_HEADERS (which apparently is not optional afer all) param in ``bottleneck/src/bn_config.py `` ``` OPTIONAL_HEADERS = [("HAVE_SSE2",...
Ok so I added the constant declaration ``const npy_DTYPE0 ai`` at lines 1316 and 1520 in ``bottleneck/src/reduce_template.c`` and the build finished successfuly.