I have an example and want to apply the function mvnpdf(x,mu,sigma) for some precalculated mu and sigma. Due to the dimensionality it will always put out 0. The solution me and friend found was applying mvn(vpa(x),mu,sigma) instead to increase precision. BUT for me the function then returned an error message regarding the function bsxfun asking for numerical arrays (In version 2017b & 2019b). For my friend it worked out without a problem. I could even solve the problem by exchanging the code line with the appropriate function application.
Why is matlab using bsxfun instead of the application of the function itself?