⁴⁶The diﬀerences between this function and the one required by MIPS is that (1) this one represents a single set of equality constraints or inequality constraints, not both, (2) it returns in dg the $m × n$ Jacobian (for $m$ constraints and $n$ variables), whereas MIPS expects the $n× m$ transpose, (3) this function allows arbitrary additional parameters, and (4) x can be a cell array of sub-vectors of optimization variable $ˆx$ (see varsets in Table 7-1).