-
we currently default to the reparameterization gradient if the `Variational` class implements `reparam`
however, if the `Inference` class does not support reparameterization gradients (e.g. `KLpq`) t…
-
Update from Dimitri, Oct 2012: after talking with many developers and users, we should definitely make the inversion tools in SPECFEM3D more standard and more user friendly. Even if there are many opt…
-
Looks like the drop changes in 0xc change the block/br type system significantly. I've been trying to figure out what's going on from the new spec tests, but can't seem to. Is this documented somewher…
-
After fitting a `sm.sandbox.gam.AdditiveModel` on a medium-sized dataset (8000 samples x 47 features), I'm getting some really funny results. All my `results.offset` and `results.mu` items are on the …
-
The stepsize in fmin_bfgs doesn't look robust when the gradient is large, and results in convergence failures.
I tried a few examples with fmin_l_bfgs_b and didn't have any convergence problems.
(asi…
-
I'm trying to sample from a complex hierarchical model with lots of variables. I'm thinking of applying ADVI to sample from the posterior because the model has potentially many variables. One of the v…
-
some code skeleton
-
Thanks @suiji for Rborist code. If I run it with 100 trees as in https://github.com/szilard/benchm-ml/tree/master/z-other-tools (on 32 core box) I get:
Time: 87 sec
AUC: 66.43
Something is wrong, the…
-
I think NLopt is probably the right choice. Here is a note on this from @hongkai-dai:
I did a quick search, and it seems IPOPT is a good choice. Both IPOPT and and NLopt are free and open source. IP…
-
**Initial description**
The objective of this project is to develop a novel closed form Perspective n Point (PnP) pose estimation algorithm for a monocular camera. A variety of techniques have been de…