Charlie Vanaret
Charlie Vanaret
Support for L-BFGS. Limited memory quasi-Newton Hessian approximations will be computed based on the [compact QN representation](https://en.wikipedia.org/wiki/Compact_quasi-Newton_representation) (see *Numerical optimization* by Nocedal & Wright, pp 181-184). The availability of quasi-Newton...
Create an `InequalityHandlingMethod` per phase in `FeasibilityRestoration`: - this allows a precise allocation of memory for the optimality problem and the feasibility problem; - this fixes an allocation bug in...
Interior-point methods: handle the slack variables in `PrimalDualInteriorPointProblem` instead of reformulating the `Model` into a `HomogeneousEqualityConstrainedModel`. Several goals: - handle more structure (e.g. the slacks produce a positive definite block...
Added reoptimization capabilities to Uno, using the existing `WarmstartInformation` object: - [x] pass the `WarmstartInformation` object as a parameter to `Uno::solve()`, instead of creating it within the function; - [x]...
The goal of this PR is to upgrade from 4 to 8 ingredients: - reformulation layer: - constraint relaxation strategy - inequality handling method - subproblem layer: - local model...
Here's the log for the `ipopt` preset on this [unbounded LP](https://github.com/jump-dev/MathOptInterface.jl/blob/7a0dd03758dfdd1e68e3c98c2635e971fac93f0b/src/Test/test_linear.jl#L1384) (MOI's `test_linear_DUAL_INFEASIBLE`). Although we reach feasible iterates (infeasibility in [1e-17, 1e-10]), the filter method allows increases of the infeasibility...
Bugfix: allow termination with `Unbounded` status only when primal feasible
Exceptions are thrown in several recoverable situations during the optimization process: - the functions of the optimization problem cannot be evaluated; - iterative regularization doesn't seem to succeed; - the...
A Minotaur/Uno interface is being developed and the root node was successfully solved. We now need a way to support reoptimization e.g. when the variable bounds change. I think it...