Only compute Jacobian in RLM if last step success (#303)

Co-authored-by: Johannes Terblanche <Affie@users.noreply.github.com>

Author: Affie

src/plans/nonlinear_least_squares_plan.jl

@@ -192,6 +192,7 @@ mutable struct LevenbergMarquardtState{
     damping_term_min::Tparams
     β::Tparams
     expect_zero_residual::Bool
+    last_step_successful::Bool
     function LevenbergMarquardtState(
         M::AbstractManifold,
         p::P,
@@ -244,6 +245,7 @@ mutable struct LevenbergMarquardtState{
             damping_term_min,
             β,
             expect_zero_residual,
+            true,
         )
     end
 end

src/solvers/LevenbergMarquardt.jl

@@ -264,7 +264,10 @@ function step_solver!(
     M = get_manifold(dmp)
     nlso = get_objective(dmp)
     basis_ox = _maybe_get_basis(M, lms.p, nlso.jacobian_tangent_basis)
-    get_jacobian!(dmp, lms.jacF, lms.p, basis_ox)
+    # a new Jacobian is only  needed if the last step was successful
+    if lms.last_step_successful
+        get_jacobian!(dmp, lms.jacF, lms.p, basis_ox)
+    end
     λk = lms.damping_term * norm(lms.residual_values)^2
 
     JJ = transpose(lms.jacF) * lms.jacF + λk * I
@@ -292,8 +295,10 @@ function step_solver!(
         if lms.expect_zero_residual
             lms.damping_term = max(lms.damping_term_min, lms.damping_term / lms.β)
         end
+        lms.last_step_successful = true
     else
         lms.damping_term *= lms.β
+        lms.last_step_successful = false
     end
     return lms
 end