I’ve come up with what appears to be a good lattice generator, which has demonstrable (and optimizable) low discrepancy for very high dimensions. This is as yet unpublished and lacks a proof, extract and document follows…
Since the early work of Zaremba [1] in developing good lattice rules using ratios of successive terms of the Fibonacci series, there have been numerous applications to numerical integration, particularly with regards to spherical codes. Ratios of succesive elements of the Fibonacci sequence give rational approximations to the golden ratio…
Posted by bbrouwer
Posted by bbrouwer