Hello, The new MOI.VectorNonlinearOracle support providing the Hessian of the Langrangian function, defined as, according to MOI.jl doc: \sum_i \mu_i \nabla^2 f_i(\mathbf{x}) but the Langrangian typically also include the objective function: L( \mathbf{x},\boldsymbol{\mu}) = J(\mathbf{x}) + \…

[image] odow: and then compute a single Hessian of lagrange, in pseudo-code: eval_hessian_lagrangian(ret, x, u) = ret .= vec(DI.hessian(x -> lagrange(x, u))) You’d want to re-use any work for different x and u, but I don’t know how to do that in DI. And you want sparsity if it exists. So, f…

Computing Hessian of the Langrangian efficiently with DI.jl

Specific Domains Optimization (Mathematical)

odow October 30, 2025, 10:51pm 11

Providing (1, 2) is enough

1 Like

Topic		Replies	Views
Using a sparse Hessian in Optim.jl Optimization (Mathematical)	21	295	October 29, 2024
Hessian preparation using DifferentationInterface and Mooncake throws a stackoverflow error Specific Domains question	9	279	February 17, 2025
Evaluating the Hessian of the Lagrangian efficiently Performance modelingtoolkit	14	2059	January 27, 2022
JuMP interface access to automatic differentiation (ReverseDiffSparse) Optimization (Mathematical) jump , differentiation	5	2707	February 15, 2018
Lagrange multipliers of a `VectorNonlinearOracle` after solve Optimization (Mathematical) jump , mathoptinterface	1	81	November 30, 2025

Computing Hessian of the Langrangian efficiently with DI.jl

Related topics