BSSUnfold

Added

• Mystic-based unfolding — new unfold_mystic() method using the mystic constrained-optimization framework. Minimizes ||A·x − b||² + α·||x||_norm with x ≥ 0 via a quadratic penalty. Supports norm (L1/L2), smoothness constraints (order 1/2), multiple mystic solvers (fmin, fmin_powell, diffev, diffev2) and all regularization selection methods (manual/cosine/lcurve/gcv/dp).
  • New file: core/unfold_mystic.py (solve_mystic + unfold_mystic)
  • Optional dependency group: bssunfold[mystic] (mystic>=0.4.5)
  • Registered in unfold_combined() pipelines as 'mystic'
  • 24 new tests in tests/test_mystic.py
• Static (bandit, pip-audit) and dynamic (DynaPyt) security analysis.
• SMT-based unfolding — new unfold_smt() method, a port of the Haskell/SBV linearEqSolver backed by the optional Z3 solver. Minimizes ||A·x − b||₁ and then the total fluence Σx over the non-negative orthant using Z3's optimizer, with exact solvers for integer and rational systems.
  • New file: core/unfold_smt.py (solve_integer_linear_eqs, solve_integer_linear_eqs_all, solve_rational_linear_eqs, solve_rational_linear_eqs_all, solve_smt, unfold_smt)
  • New Detector.unfold_smt() method
  • Registered in unfold_combined() pipelines as 'smt'
  • Optional dependency group: bssunfold[smt] (z3-solver>=4.13.0)
  • New tests in tests/test_smt.py
• Genetic / meta-heuristic unfolding — new unfold_genetic() method using population-based meta-heuristic algorithms from MEALPY. Minimizes ||A·x − b||²/||b||² + α·||x||_norm with x ≥ 0 and optional second-difference smoothing and Shannon-entropy terms, following the PSO (Shahabinejad & Sohrabpour 2017), GA (Suman & Sarkar 2012) and entropy-based (Woo et al. 2019) unfolding works. No initial spectrum is required (random population initialization).
  • New file: core/unfold_genetic.py (solve_genetic + unfold_genetic)
  • 8 MEALPY solvers: pso (chaotic PSO, default), ga, de, es, ep, abc, gwo, cmaes
  • New Detector.unfold_genetic() method
  • Registered in unfold_combined() pipelines as 'genetic'
  • Optional dependency group: bssunfold[mealpy] (mealpy>=3.0.2)
  • 30 new tests in tests/test_genetic.py
  • scip and cplex
  • Compressive Sensing (CS) unfolding — new unfold_cs() method based on compressive sensing. The spectrum is represented sparsely in a learned dictionary (x = D @ alpha), the dictionary is learned with K-SVD, sparse coding is performed with OMP, and reconstruction is done with the SL0 algorithm. Well suited for the highly underdetermined problem where the number of energy groups greatly exceeds the number of detector readings.
  • New file: core/unfold_cs.py (solve_omp, solve_ksvd, solve_sl0, solve_cs, unfold_cs)
  • New Detector.unfold_cs() method
  • No extra dependencies (pure NumPy)
  • 21 new tests in tests/test_cs.py
  • New example notebook: examples/23-CS.ipynb

Changed

• numba promoted to a core dependency (the bssunfold[numba] extra is kept for compatibility).
• Test-suite coverage gate raised to 98.3% (was 91.9%): numba JIT bodies are excluded from coverage via # pragma: no cover (compiled to LLVM, never run as CPython bytecode) and targeted branch tests added in tests/test_coverage.py and tests/test_coverage_boost.py.