sleipnirgroup-jormungandr 0.6.7.dev30__tar.gz → 0.6.7.dev32__tar.gz

{sleipnirgroup_jormungandr-0.6.7.dev30 → sleipnirgroup_jormungandr-0.6.7.dev32}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sleipnirgroup-jormungandr
-Version: 0.6.7.dev30
+Version: 0.6.7.dev32
 Summary: Reverse mode autodiff library and NLP solver DSL
 License-Expression: BSD-3-Clause
 License-File: LICENSE.txt

{sleipnirgroup_jormungandr-0.6.7.dev30 → sleipnirgroup_jormungandr-0.6.7.dev32}/include/sleipnir/optimization/solver/util/feasibility_restoration.hpp

@@ -634,12 +634,22 @@ ExitStatus feasibility_restoration(
                   fr_x, fr_s, fr_y, fr_z, fr_μ, iterations);
 
   x = fr_x.segment(0, x.rows());
-  s = fr_s.segment(0, s.rows());
 
   if (status == ExitStatus::CALLBACK_REQUESTED_STOP) {
-    auto g = matrices.g(x);
-    auto A_e = matrices.A_e(x);
-    auto A_i = matrices.A_i(x);
+    // Re-center the slacks on the original constraints cᵢ(x) if they're
+    // sufficiently feasible, and on 1 otherwise.
+    //
+    // We can't use the slacks of cᵢ(x) − pᵢ + nᵢ ≥ 0 because feasibility
+    // restoration's barrier parameter fr_μ drives them toward 0 independently
+    // of the original problem's barrier parameter μ. Slacks much smaller than μ
+    // make Σ = S⁻¹Z and μS⁻¹e very large, which causes very large dual steps
+    // after returning.
+    const DenseVector c_i = matrices.c_i(x);
+    s = (c_i.array() >= Scalar(1e-2)).select(c_i, Scalar(1));
+
+    const SparseVector g = matrices.g(x);
+    const SparseMatrix A_e = matrices.A_e(x);
+    const SparseMatrix A_i = matrices.A_i(x);
 
     auto [y_estimate, z_estimate] =
         lagrange_multiplier_estimate(g, A_e, A_i, s, μ);
@@ -654,6 +664,9 @@ ExitStatus feasibility_restoration(
     // infeasibility anywhere else risks false positives (e.g., a
     // point-in-time test can reject iterates the solver would otherwise
     // escape). See section 3.3, p. 14 of [2].
+
+    s = fr_s.segment(0, s.rows());
+
     DenseVector c_e = matrices.c_e(x);
     if (matrices.scaling.c_e.size() > 0) {
       c_e = matrices.scaling.c_e.cwiseInverse().cwiseProduct(c_e);
@@ -688,6 +701,7 @@ ExitStatus feasibility_restoration(
 
     return ExitStatus::FEASIBILITY_RESTORATION_FAILED;
   } else {
+    s = fr_s.segment(0, s.rows());
     return ExitStatus::FEASIBILITY_RESTORATION_FAILED;
   }
 }

{sleipnirgroup_jormungandr-0.6.7.dev30 → sleipnirgroup_jormungandr-0.6.7.dev32}/python/src/sleipnir/__init__.py

@@ -2,4 +2,4 @@
 
 from ._sleipnir import *
 
-__version__ = "0.6.7.dev30"
+__version__ = "0.6.7.dev32"

{sleipnirgroup_jormungandr-0.6.7.dev30 → sleipnirgroup_jormungandr-0.6.7.dev32}/python/test/optimization/nonlinear_problem_test.py

@@ -124,7 +124,9 @@ def test_conflicting_bounds():
 
 
 def test_wachter_and_biegler_line_search_failure():
-    # See example 19.2 of [1]
+    # Example 19.2 of [1] exercises a deficiency of line search IPMs that must
+    # be overcome by feasibility restoration. Depending on the specific
+    # trajectory taken, either success or local infeasibility are expected.
     #
     # [1] Nocedal, J. and Wright, S. "Numerical Optimization", 2nd. ed., Ch. 19.
     #     Springer, 2006.
@@ -147,8 +149,14 @@ def test_wachter_and_biegler_line_search_failure():
     assert problem.equality_constraint_type() == ExpressionType.QUADRATIC
     assert problem.inequality_constraint_type() == ExpressionType.LINEAR
 
-    assert problem.solve(diagnostics=True) == ExitStatus.SUCCESS
-
-    assert x.value() == pytest.approx(1.0, abs=1e-6)
-    assert s1.value() == pytest.approx(0.0, abs=1e-6)
-    assert s2.value() == pytest.approx(0.5, abs=1e-6)
+    status = problem.solve(diagnostics=True)
+    assert status in (ExitStatus.SUCCESS, ExitStatus.LOCALLY_INFEASIBLE)
+
+    if status == ExitStatus.SUCCESS:
+        assert x.value() == pytest.approx(1.0, abs=1e-6)
+        assert s1.value() == pytest.approx(0.0, abs=1e-6)
+        assert s2.value() == pytest.approx(0.5, abs=1e-6)
+    else:
+        assert x.value() == pytest.approx(-1.0, abs=1e-6)
+        assert s1.value() == pytest.approx(0.0, abs=1e-6)
+        assert s2.value() == pytest.approx(0.0, abs=3e-5)