Hafnian

Summary of changes

New features

- Added two new sampling functions in `hafnian.sample`, allowing efficient sampling from classical Gaussian states:

* `hafnian_sample_classical_state`
* `torontonian_sample_classical_state`

- Added functions to calculate the probability amplitudes directly for pure states:

* `pure_state_amplitude`
* `state_vector`

- Added utility functions to check the validity of covariance matrices:

* `is_valid_cov`
* `is_pure_cov`
* `is_classical_cov`

Improvements

* All functions in the `quantum` package now take as arguments xp-covariance matrices and vectors of means for consistency.

Contributors

This release contains contributions from (in alphabetical order):

Nicolas Quesada

Summary of changes

New features

- Added a new sampling submodule `hafnian.sample`, allowing sampling from the underlying hafnian/Torontonian distribution of graphs/Gaussian states

- Documentation overhaul: now contains some of the best and most up-to-date information about hafnians, loop hafnians, and Torontonians

- C++ library has been significantly improved, tested, and refactored

Improvements

- Ported the `hafnian_approx.F90` Fortran file to C++

- The Torontonian function is now parallelized via OpenMP

- Tests and the C++ header library have been refactored

- Addition of new C++ tests using Googletest

- C++ library is now documented via Doxygen; this is integrated directly into Sphinx

Contributors

This release contains contributions from (in alphabetical order):

Brajesh Gupt, Josh Izaac, Nicolas Quesada

Bug fixes

* Solve type conflict preventing the pip wheel to build correctly (see 4).

Contributors

emaballarin

New features

* Odd-dimensioned matrices no longer raise an error, as the loop hafnian is well defined for matrices with odd dimensions. The non-loop hafnian will now return `0` for matrices with odd-dimensions.