From 3cf342ab31a9876f0282962154d3f2b4e6bf6e59 Mon Sep 17 00:00:00 2001 From: ChrisRackauckas-Claude Date: Mon, 9 Feb 2026 14:26:32 -0500 Subject: [PATCH] Fix flaky Lorenz System test by setting deterministic random seed The Lorenz System parameter estimation test in NNPDE2 was failing intermittently because the neural network weight initialization used a non-deterministic random seed, causing BFGS optimization to sometimes converge to bad local minima. Add Random.seed!(100) for reproducibility, consistent with other test items in this file. Co-Authored-By: Chris Rackauckas Co-Authored-By: Claude Opus 4.6 --- test/additional_loss_tests.jl | 2 ++ 1 file changed, 2 insertions(+) diff --git a/test/additional_loss_tests.jl b/test/additional_loss_tests.jl index 5f39ad50b..c491b1f47 100644 --- a/test/additional_loss_tests.jl +++ b/test/additional_loss_tests.jl @@ -112,6 +112,8 @@ end import DomainSets: Interval, infimum, supremum using OptimizationOptimJL: BFGS + Random.seed!(100) + @parameters t, σ_, β, ρ @variables x(..), y(..), z(..) Dt = Differential(t)