Johan Löfberg

Currently not really defined what it returns

Robustify fails when using norm

1

``` zeta = sdpvar(2,1); w = sdpvar(2,1); F = [uncertain(zeta); norm(zeta)

Can be done by introducing auxilliary continuous model x*z ``` binvar x sdpvar z binmodel(x*z^2,[0

Can happen if dualization is done automatically due to the -1 flag as implemented in issue #80

Could perhaps be a general feature for all operators lacking a differential operator

Can improve performance significantly if we skip this normalization in yalmip.m

Is it possible to use cplex iis to detect the problematic ocnstraints in a model?

``` sdpvar x y ncvar z replace(x*y,x,z) ```

Overload the struct functionalities etc for back-wards compatibility, but make it an object which can be manipulated further (optimize()).value etc