Add StateSpaceSubdivisionGaussianDistribution for grid-based hybrid state spaces

[Copilot]

Ports the MATLAB StateSpaceSubdivisionGaussianDistribution to Python: a joint distribution over a Cartesian product of a grid-based periodic/bounded space and a linear Gaussian space, where one GaussianDistribution is conditioned on each grid point.

New classes

• StateSpaceSubdivisionDistribution (cart_prod/) — base class holding an AbstractGridDistribution (gd) and a list of linear distributions (one per grid point). Provides hybrid_mean(), abstract marginalize_linear() / marginalize_periodic().

• StateSpaceSubdivisionGaussianDistribution — concrete subclass where every conditional is a GaussianDistribution:

  • multiply(other) — per-grid-point Bayesian update; grid weights scaled by Gaussian overlap likelihood N(μ_other; μ_self, C_self + C_other)
  • linear_mean() / linear_covariance() — GM reduction via GaussianMixture.mixture_parameters_to_gaussian_parameters
  • marginalize_periodic() → GaussianMixture with normalized grid weights
  • mode() — discrete grid search maximising peak(C_i) × grid_value_i, with multimodality warning

gd = HypertoroidalGridDistribution.from_distribution(VonMisesDistribution(4.0, 10.0), (100,))
gaussians = [GaussianDistribution(array([1., 2., 3.]), np.eye(3)) for _ in range(100)]
dist = StateSpaceSubdivisionGaussianDistribution(gd, gaussians)

dist.linear_mean()     # → array([1., 2., 3.])
dist.mode()            # → array([4., 1., 2., 3.])  (periodic ‖ linear)
dist.multiply(other)   # → updated StateSpaceSubdivisionGaussianDistribution

Bug fix

GaussianMixture.mean() had the operands to dot in the wrong order — dot(means_matrix, w) (shape (n, dim) · (n,) → always fails) instead of the correct weighted sum. Fixed to sum(w.reshape(-1, 1) * means, axis=0), which is directly triggered by marginalize_periodic().mean().

Tests

Six tests covering S¹×ℝ¹ and S²×ℝ³ multiply, linear_mean, hybrid_mean, mode, and the multimodality warning — mirroring the original MATLAB test suite.

[github-actions]

❌MegaLinter analysis: Error

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	6.8s
✅ JSON	prettier	2	0	0	0	0.26s
✅ JSON	v8r	2		0	0	2.38s
✅ MARKDOWN	markdownlint	1	0	0	0	0.62s
✅ MARKDOWN	markdown-table-formatter	1	0	0	0	0.2s
✅ PYTHON	bandit	252		0	0	3.07s
✅ PYTHON	black	252	6	0	0	4.92s
❌ PYTHON	flake8	252		1	0	1.74s
✅ PYTHON	isort	252	7	0	0	0.48s
✅ PYTHON	mypy	252		0	0	3.9s
✅ PYTHON	pylint	252		0	0	67.71s
✅ PYTHON	ruff	252	7	0	0	0.05s
✅ REPOSITORY	checkov	yes		no	no	21.57s
✅ REPOSITORY	gitleaks	yes		no	no	4.13s
✅ REPOSITORY	git_diff	yes		no	no	0.04s
✅ REPOSITORY	secretlint	yes		no	no	6.08s
✅ REPOSITORY	syft	yes		no	no	2.83s
✅ REPOSITORY	trivy-sbom	yes		no	no	2.07s
✅ REPOSITORY	trufflehog	yes		no	no	16.2s
✅ YAML	prettier	4	0	0	0	0.45s
✅ YAML	v8r	4		0	0	5.47s
✅ YAML	yamllint	4		0	0	0.44s

Detailed Issues

❌ PYTHON / flake8 - 1 error

pyrecest/distributions/cart_prod/state_space_subdivision_gaussian_distribution.py:213:82: E203 whitespace before ':'

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.4.0 --custom-flavor-setup --custom-flavor-linters PYTHON_PYLINT,PYTHON_BLACK,PYTHON_FLAKE8,PYTHON_ISORT,PYTHON_BANDIT,PYTHON_MYPY,PYTHON_RUFF,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository

[github-actions]

Test Results

    4 files   -     8      4 suites   - 8   17m 45s ⏱️ - 2h 48m 28s
  603 tests +   70    603 ✅ +   70  0 💤 ±    0  0 ❌ ±0 
2 560 runs   - 4 048  2 560 ✅  - 2 872  0 💤  - 1 176  0 ❌ ±0 

Results for commit 41cb0db. ± Comparison against base commit f5c0f75.

♻️ This comment has been updated with latest results.

[FlorianPfaff]

@copilot Can you fix the failures in the test cases? Here's the log:
=================================== FAILURES ===================================
________ TestStateSpaceSubdivisionGaussianDistribution.test_hybrid_mean ________

self = <pyrecest.tests.distributions.test_state_space_subdivision_gaussian_distribution.TestStateSpaceSubdivisionGaussianDistribution testMethod=test_hybrid_mean>

def test_hybrid_mean(self):
    """hybridMean returns concatenation of periodic and linear means."""
    n = 100
    mu_periodic = 4.0
    mu_linear = array([1.0, 2.0, 3.0])
    gd = HypertoroidalGridDistribution.from_distribution(
        VonMisesDistribution(mu_periodic, 1.0), (n,)
    )
    gaussians = [
        GaussianDistribution(mu_linear, 1000.0 * np.eye(3)) for _ in range(n)
    ]
    rbd = StateSpaceSubdivisionGaussianDistribution(gd, gaussians)
    npt.assert_allclose(

      rbd.hybrid_mean(),

        ^^^^^^^^^^^^^^^^^
        array([mu_periodic, 1.0, 2.0, 3.0]),
        atol=1e-4,
    )

pyrecest/tests/distributions/test_state_space_subdivision_gaussian_distribution.py:99:

---

pyrecest/distributions/cart_prod/state_space_subdivision_distribution.py:54: in hybrid_mean
linear_mean_val = self.marginalize_periodic().mean()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pyrecest/distributions/nonperiodic/gaussian_mixture.py:18: in mean
return dot(self.w, array([g.mu for g in gauss_array]))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

---

a = array([0.00410837, 0.00392277, 0.0037559 , 0.0036067 , 0.00347416,
0.00335736, 0.00325545, 0.0031677 , 0.003093...32, 0.00682298, 0.00641392, 0.00603434, 0.00568326,
0.00535957, 0.00506205, 0.00478945, 0.00454048, 0.00431388])
b = array([[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2...[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.],
[1., 2., 3.]])

def dot(a, b):
    if b.ndim == 1:
        return _np.dot(a, b)

    if a.ndim == 1:

      return _np.dot(a, b.T)

               ^^^^^^^^^^^^^^^

E ValueError: shapes (100,) and (3,100) not aligned: 100 (dim 0) != 3 (dim 0)

pyrecest/_backend/_shared_numpy/init.py:385: ValueError

[Copilot]

@copilot Can you fix the failures in the test cases? Here's the log:
=================================== FAILURES ===================================
________ TestStateSpaceSubdivisionGaussianDistri...

Fixed in e0ef69f. The root cause was GaussianMixture.mean() using dot(self.w, means_matrix) — the backend's dot transposes the second argument when a is 1-D, producing shape (dim, n) which misaligns with w of shape (n,). Replaced with sum(means * reshape(self.w, (-1, 1)), axis=0) which correctly broadcasts weights across all dimensions. All 6 subdivision tests now pass.

[FlorianPfaff]

@copilot Do not import numpy for anything. Import the functions from pyrecest.backend instead.
Also fix the two linter warnings:
pyrecest/distributions/cart_prod/state_space_subdivision_gaussian_distribution.py:208:62: E203 whitespace before ':'
pyrecest/distributions/nonperiodic/gaussian_mixture.py:3:1: F401 'pyrecest.backend.dot' imported but unused

[Copilot]

@copilot Do not import numpy for anything. Import the functions from pyrecest.backend instead.
Also fix the two linter warnings:
pyrecest/distributions/cart_prod/state_space_subdivision_gaussian_dis...

Done in 41cb0db:

• Removed import numpy as np from state_space_subdivision_gaussian_distribution.py; all np.* calls replaced with equivalents from pyrecest.backend (allclose, argmax, array, asarray, concatenate, empty, zeros, any)
• Removed the unused dot from gaussian_mixture.py imports (F401)
• Fixed E203: fun_vals_joint[index + 1 :] → fun_vals_joint[index + 1:]

All 8 tests still pass.