linear regression system identification

hello everybody,i have to do a system identification task with linear regression method,but instead of using the 2_norm of the error for optimization,i have to use the inf and first norm of errors,is there anybody help me in writing the code?

Answer 1

[Matt J]

For the inf-norm, you could use FMINIMAX. For the L1-norm, you would probably have to reformulate as follows and minimize over (x,y) using LINPROG

 min sum(y)

 s.t.
  error(x)-y<=0
 -error(x)-y<=0