v1.1.1

Respect show_progress argument in main spectral observable computation functions.

Bugfix: Remove Numba caching

Remove caching as a way to temporarily workaround numba/numba#4908

v1.0.0

Included in this update:

Core Algorithm Improvements

• improves the spectral rigidity and level variance algorithms
  • algorithms now automatically determine the number of iterations needed for convergence, based on largest L value
  • both algorithms now use more consistent Monte-Carlo approaches with a clear convergence criterion
  • Kahan summation is now used for the summing/averaging of Monte-Carlo sample runs (doesn't seem to make a difference though - maybe Numba compiling it out...)
  • progress monitoring during computation of spectral observables has been simplified

API Changes / Improvements

• Eigenvalues class constructor now accepts 2D matrices as inputs, and automatically tries to use the appropriate eigenvalue computation function to extract eigenvalues
• generate_eigs function moved to Eigenvalues.generate static method
• most string literals (e.g. "goe", "gue") have been replaced with Enum variants:
  • empyricalRMT._types.MatrixKind
  • empyricalRMT.plot.PlotMode
  • empyricalRMT.smoother.SmoothMethod
• plotting no longer forces Seaborn style "darkgrid" on user
• passing in existing plt.Figure and plt.Axes objects to plotting functions should now always be possible

Code Quality

• updates Python version and library requirements
• improved type annotations, remove issues found by updated Pylance, etc.
• consistently use np.float64 almost everywhere

This update will likely be the last, except for minor bugfixes.

Preliminary Release

Preliminary release for citation purposes.