callbacks.rst

Callbacks

Applications may define functions for evaluating problem elements given a current solution. This section of the documentation details the function signatures expected for the callbacks.

eval_f

Returns the value of the objective function at the current solution x:

function eval_f(x::Vector{Float64})   # (length n) Current Solution
  # ...
  return obj_value
end

eval_g

Sets the value of the constraint functions g at the current solution x:

function eval_g(x::Vector{Float64},           # (length n) Current Solution
                  cons::Vector{Float64})      # (length m) Constraint values g(x)
  # ...
  # cons[1] = ...
  # ...
  # cons[prob.m] = ...
end

Note that the values of cons must be set "in-place", i.e. the statement cons = zeros(prob.m) musn't be done. If you do want to create a new vector and allocate it to cons use cons[:], e.g. cons[:] = zeros(prob.m).

eval_grad_f

Sets the value of the gradient of the objective function at the current solution x:

function eval_grad_f(x::Vector{Float64},          # (length n) Current Solution
                       grad::Vector{Float64})     # (length n) The gradient of the objective function
  # ...
  # grad[1] = ...
  # ...
  # grad[prob.n] = ...
end

As with eval_g, you must set the values "in-place" for eval_grad_f.

eval_jac_g

This function returns the values of the Jacobian, evaluated at the non-negative indices, based on the sparsity structure passed to KNITRO through initializeProblem. Julia is 1-based, in the sense that indexing always starts at 1 (unlike C, which starts at 0).:

function eval_jac_g(x::Vector{Float64},       # (length n) Current Solution
                      jac::Vector{Float64})   # (length nnzJ) The values of the Jacobian
  # ...
  # jac[1] = ...
  # ...
  # jac[nnzJ] = ... # where nnzJ = length(jac)
end

As for the previous two callbacks, all values must be set "in-place". See the Ipopt documentation for a further description of the sparsity format followed by Ipopt ((row,column,value) triples).

eval_h

Similar to the Jacobian, except for the Hessian of the Lagrangian. See documentation for full details of the meaning of everything.:

function eval_h(x::Vector{Float64},        # (length n) Current solution
                  lambda::Vector{Float64},   # (length n+m) Multipliers for each constraint
                  sigma::Float64,            # Lagrangian multiplier for objective
                  hess::Vector{Float64})     # (length nnzH) The values of the Hessian
  # ...
  # hess[1] = ...
  # ...
  # hess[nnzH] = ... # where nnzH = length(hess)
end

eval_hv

Computes the Hessian-of-the-Lagrangian-vector product, storing the result in the vector hv.:

function eval_hv(x::Vector{Float64},        # (length n) Current solution
                   lambda::Vector{Float64},   # (length n+m) Multipliers for each constraint
                   sigma::Float64,            # Lagrangian multiplier for objective
                   hess::Vector{Float64})     # (length n) Hessian-of-the-Lagrangian-vector product
  # ...
  # hv[1] = ...
  # ...
  # hv[end] = ...
end