search

AusterweilLab / MPBNP

An OpenCL-based toolkit for performing inference for Bayesian Nonparametric Models
MIT License
11 stars 1 forks source link

Models to Estimating Semantic Network #5

Open hyiltiz opened 8 years ago

hyiltiz commented 8 years ago

I am sure this is not the place to ask about how is your model on estimating semantic network, but I think it is better than an email. I believe it is still a work in progress. I am quite interested in the collection (set) of categories that are used to construct the semantic network. Could you refer to me your methods, or program of your model?

BTW, maybe Julia is way quicker than Python and suitable to do OpenCL.

tqian86 commented 8 years ago

Hi hyiltiz, thanks for the question. I'll defer to Joe to answer any question regarding semantic networks. Regarding Python vs Julia, we are aiming for the common denominator in the research community. Python is much more widely used than Julia. Also, the Python side of code in MPBNP is heavily vectorized and implemented via Numpy, which hooks to BLAS. I don't think Julia is somehow magically faster than a BLAS library, right? ;-) The so-called speed up in Julia usually applies to the limited cases of arithmetic loops - which have been carefully avoided as much as possible in the development of MPBNP.

hyiltiz commented 8 years ago

Yeah, BLAS rocks. I am really interested in the semantic networks and thank you!

On Fri, Nov 13, 2015, 08:22 Ting Qian notifications@github.com wrote:

Hi hyiltiz, thanks for the question. I'll defer to Joe to answer any question regarding semantic networks. Regarding Python vs Julia, we are aiming for the common denominator in the research community. Python is much more widely used than Julia. Also, the Python side of code in MPBNP is heavily vectorized and implemented via Numpy, which hooks to BLAS. I don't think Julia is somehow magically faster than a BLAS library, right? ;-) The so-called speed up in Julia usually applies to the limited cases of arithmetic loops - which have been carefully avoided as much as possible in the development of MPBNP.

— Reply to this email directly or view it on GitHub https://github.com/AusterweilLab/MPBNP/issues/5#issuecomment-156430445.

Sent from Gmail Mobile.