Continuing the implementation of the associated paper at NeurIPS '19 on very fast and exact DPP sampling with intermediate sampling and Nystrom approximation.
Added docstrings for most new functions.
The code is ready to be tested e.g. using a shim likelihood function such as
def linear_likelihood(X, Y=None):
if Y is None:
return X.dot(Y.T)
else:
return X.dot(X.T)
Continuing the implementation of the associated paper at NeurIPS '19 on very fast and exact DPP sampling with intermediate sampling and Nystrom approximation.
Added docstrings for most new functions.
The code is ready to be tested e.g. using a shim likelihood function such as