search

probcomp / Venturecxx

Primary implementation of the Venture probabilistic programming system
http://probcomp.csail.mit.edu/venture/
GNU General Public License v3.0
28 stars 6 forks source link

20160830 schaechtle gpgradient b #628

Closed Schaechtle closed 8 years ago

Schaechtle commented 8 years ago

Linear kernel now takes multivariate input.

Computing the linear kernel as the inner product: if X = \R^n, then we can define a 'linear' covariance kernel L: X x X ---> \R by L(x, y) = <x, y>, which reduces to the usual linear kernel when n = 1.

In addition to the covariance.py, gp.py was changed to allow for the new type signature. Furthermore, _cov_maker(...) was changed to _cov_grad_maker(...) to accommodate the gradient for this kernel.

Finally a test wast added to test/conformance/sps/test_gp.py to test this new kernel with multivariate input.

The changes were tested by running:

   $  ./check.sh  -c crashes.cfg test/conformance/sps/test_gp.py

No errors occurred.

axch commented 8 years ago

Closing as this was effectively (but not literally) merged a few days ago.

riastradh-probcomp commented 8 years ago

Specifically, supplanted by 1b04db069ada1abb9a357675cffc3e19280af498.