Ok, for example, here g and lam are both column vectors and dg (for compatibility with fmincon) is a matrix whose columns correspond to the rows of g.
Ray > On Aug 30, 2019, at 11:17 AM, Jubeyer Rahman <[email protected]> wrote: > > Well I am referring to line 390 and 392 in mips.m. > > Regards, > Jubeyer > > On Fri, Aug 30, 2019 at 10:58 AM Ray Zimmerman <[email protected] > <mailto:[email protected]>> wrote: > The mips() function requires that you pass in function handles for functions > that evaluate the objective and it’s gradient (f_fcn), and the constraints > and their gradients (gh_fcn) or Jacobian, and Hessian (hess_fcn). So, mips > itself does not compute these quantities (e.g. via finite differences), but > it uses the functions it is provided. > > Regarding the transpose matrices, I’m not sure where you are referring to > (maybe provide a specific file and line number), but MIPS was designed to be > as compatible as possible with the way derivative and Hessian functions were > defined for fmincon (part of MATLAB’s Optimization Toolbox) and I believe > that required a transpose somewhere. > > Best, > > Ray > > > > >> On Aug 29, 2019, at 4:55 PM, Jubeyer Rahman <[email protected] >> <mailto:[email protected]>> wrote: >> >> Hi, >> Is there any Jacobian calculation taking place in mips.m? If not, what >> function can I use to calculate the Jacobian matrix inside the mips where >> Lxx is calculated (hessian)? >> >> Another question is, while taking the derivative to produce Lx from L ; why >> the transpose matrices become regular, say for example, lam' becomes lam, >> etc.? >> >> -Jubeyer >> >
