revrand
revrand copied to clipboard
NICTA
A library of scalable Bayesian generalised linear models with fancy features
Hey guys, searching for nlopt Python tutorials I found this wrapper which at some time belonged to the revrand project apparently: https://www.earthbyte.org/webdav/ftp/Data_Collections/Dutkiewicz_etal_2016_G3/revrand-master/revrand/optimize/nlopt_wrap.py In a nutshell, the wrapper enables to call...
Dear Project developers: Thank you for providing such a great tool. I am currently testing the "A la Carte" method proposed by Zichao Yang et al. And I notice that...
Otherwise we get potentially massive memory usage, i.e. using the SLM on SARCOS, ARD on 21 dimensions uses too much memory. Make these sorts of lines return generators, and fix...
Line profiling: - [x] GLM - [ ] SLM - [ ] SGD - [ ] basis functions (especially FastFood) Memory profiling: - [ ] GLM - [ ] SLM...
There doesn't seem to be any reason why we couldn't make an RFF or ORF-based Gaussian spectral mixture basis, and it would probably be faster for small to moderate input...
Orthogonal Random Features (https://arxiv.org/abs/1610.09072) - [x] RBF ORF - [ ] RBF SORF Neural Net - [ ] Random Arc-cosine (see #128) Random dot product kernels: - [ ] http://www.jmlr.org/proceedings/papers/v22/kar12/kar12.pdf...
We currently zero-pad `dPhi` in the `BasisCat` object for updating hyperparameters only depending on one basis. Would it be more efficient to slice instead? Maybe look into this with #106.
Look into: http://arxiv.org/pdf/1312.6055v3.pdf
