datamicroscopes / common Goto Github PK

if a system protobuf is found, cmake prefers to link to it instead of the one found in anaconda. this has a very high chance of not working (since the system ones are generally not compiled against libc++ and might be older)

if gcc-4.2 is found on the path, then distutils's setup() will try to use it for compilation. this is problematic for us-- we need to figure out how to tell distutils otherwise; it doesn't seem to respect CXX/CC in the environment.

I think this is the reason why I can't install

Hey Tim/Eric ... First, thanks for creating this library. I'd like to implement some Bayesian non-parametric models all related to time series and hidden markov models. These are the references if you are interested at all.

http://arxiv.org/abs/1308.4747
http://arxiv.org/abs/1505.01164
https://www.stat.washington.edu/~ebfox/publications/HDPSLDS_journal_v5_posted.pdf

I've implemented Bayesian models before, but they were too slow to be practical. I wondered if there was some basic overview of the base datamicroscopes library and how it used by the other models or even just some basic advice on the speeding up the inference.

this is a numpy bug: numpy/numpy#4846

but for the time being, this means that if you have a multidimensional likelihood model (e.g. NIW) then you can't have missing data.