test: add initial test of optimizer (#167)

* test: add initial test of optimizer

* style: switch to local imports and plural folders

* style: add README to sNMF test

---------

Co-authored-by: John Halloran <[email protected]>

Author: john-halloran

pyproject.toml

@@ -81,3 +81,9 @@ exclude = '''
   | tests/data
 )/
 '''
+
+[tool.pytest.ini_options]
+markers = [
+    "slow: marks tests as slow (deselect with '-m \"not slow\"')",
+]
+testpaths = ["tests"]

src/diffpy/snmf/__init__.py

@@ -18,6 +18,8 @@
 # package version
 from diffpy.snmf.version import __version__
 
+__all__ = ["__version__", "SNMFOptimizer"]
+
 # silence the pyflakes syntax checker
 assert __version__ or True
 

src/diffpy/snmf/main.py

@@ -1,11 +1,12 @@
 import numpy as np
-from snmf_class import SNMFOptimizer
+
+from diffpy.snmf.snmf_class import SNMFOptimizer
 
 # Example input files (not provided)
-init_components_file = np.loadtxt("input/init_components.txt", dtype=float)
-source_matrix_file = np.loadtxt("input/source_matrix.txt", dtype=float)
-init_stretch_file = np.loadtxt("input/init_stretch.txt", dtype=float)
-init_weights_file = np.loadtxt("input/init_weights.txt", dtype=float)
+init_components_file = np.loadtxt("inputs/init_components.txt", dtype=float)
+source_matrix_file = np.loadtxt("inputs/source_matrix.txt", dtype=float)
+init_stretch_file = np.loadtxt("inputs/init_stretch.txt", dtype=float)
+init_weights_file = np.loadtxt("inputs/init_weights.txt", dtype=float)
 
 my_model = SNMFOptimizer(
     source_matrix=source_matrix_file,

src/diffpy/snmf/snmf_class.py

@@ -1,9 +1,10 @@
 import cvxpy as cp
 import numpy as np
-from plotter import SNMFPlotter
 from scipy.optimize import minimize
 from scipy.sparse import coo_matrix, diags
 
+from diffpy.snmf.plotter import SNMFPlotter
+
 
 class SNMFOptimizer:
     """An implementation of stretched NMF (sNMF), including sparse stretched NMF.
@@ -625,7 +626,7 @@ def update_stretch(self):
         Updates matrix A using constrained optimization (equivalent to fmincon in MATLAB).
         """
 
-        # Flatten A for compatibility with the optimizer (since SciPy expects 1D input)
+        # Flatten A for compatibility with the optimizer (since SciPy expects 1D inputs)
         stretch_flat_initial = self.stretch.flatten()
 
         # Define the optimization function

tests/inputs/test_snmf_optimizer/README

@@ -0,0 +1,21 @@
+This test is a whole-program test of the SNMF Optimizer. It runs the optimization on known data with fixed, trusted settings, then asserts
+whether the objective function is below a certain value after exactly a certain number of iterations.
+It is a blunt-force instrument that can tell if a PR has completely broken things, but it will not catch regressions, and its method
+of using objective function only can inadverdently validate a degenerate result. Use with caution.
+
+This test uses data published in:
+Stretched Non-negative Matrix Factorization[J].
+Section 5.4 XRD
+Tab. 7, Fig. 8 & 9
+Gu R, Rakita Y, Lan L, et al.
+Stretched Non-negative Matrix Factorization[J].
+arXiv preprint arXiv:2311.15173, 2023.
+and later republished as:
+Gu, R., Rakita, Y., Lan, L. et al. Stretched non-negative matrix factorization.
+npj Comput Mater 10, 193 (2024). https://doi.org/10.1038/s41524-024-01377-5
+Data description:
+"A stoichiometric mixture of 2:1 YOCl (>98% tetragonal phase) and MgMn2O4 (spinel phase) was uniformly mixed and sealed in a quartz capillary. It was then heated in a gradient furnance, meaning that each location on the quartz tube had a different temperature31. The absolute temperatures at each point along the sample were calibrated from the lattice expansion of a known calibration material, Ni. The data went from a low temperature of 368∘C to a highest temperature of 668∘C with a total of 20 individual temperature points. Using ‘pyFAI’32, the collected 2D diffraction patterns were then cleaned by masking the beam-stop and over-bright/dead pixels, followed by an azimuthal integration to gain 1D PXRD patterns. The 1D PXRD data was then used as inputs to the different NMF algorithms."
+The data was originally sourced from:
+O’Nolan D, Huang G, Kamm GE, et al (2020)
+A thermal-gradient approach to variable-temperature measurements resolved in space.
+Journal of Applied Crystallography 53(3):662–670.