Support for Implicit Expansion

[sehyoun]

Matlab introduce implicit arithmetic expansion with R2016b (read here) Current implementation does not handle the implicit expansion. It will be updated in the future.

In the mean time, the problem can be bypassed using repmat. For example, let x be a (n x m) matrix and y a (n x 1) vector. With the new implicit expansion, you should be able to call

z = x + y;

However, this is equivalent to calling

z = x + repmat(y,1,m);

The first expression would throw an error with this implementation of auto diff, but latter would not. Hence, one should should the second expression while the problem is being handled.