Inflation v2.0.3

This update brings:

• Support for symmetries in the target distributions.
• Initial (and experimental) support for quantum theory based only on real numbers.

Also, we have the following improvements:

• The arguments for InflationSDP.build_columns can be now directly passed onto InflationSDP.generate_relaxation. This means that one can now build directly SDPs for restricted generating sets just by calling, e.g., InflationSDP.generate_relaxation("local3", max_monomial_length=2).
• The computation of dual LPs is now handled automatically by MOSEK. This improves the reliability of the code (and reduces the maintenance effort). The analogous for SDPs will come soon.
• The error messages for unsupported scenarios (namely, structures where a parent and a child node do not share a latent common cause, and structures with intermediate latents that have observable parents) are now more informative.

As well as minor fixes. On particularly useful one is that now the output status of LPs and SDPs is now the same.

Inflation v2.0.2

This is a release that, importantly, implements a bigfix when generating and identifying physical monomials in scenarios with intermediate latents. The changes are

• Compatibility fixes and improvements: the uses of SciPy’s coo_matrix (to be deprecated) have been substituted by coo_array. Also, the library does not need networkx any more.

• Bug fixes in the generation of physical monomials in structures with intermediate latents.

• Memory and runtime improvements.

Inflation v2.0.1

This is a minor release that implements several improvements in compatibility, memory and runtime. The changes are

• Support for NumPy 2.

• Small bugfixes.

• Memory and runtime improvements by better handling of sparse objects.

Inflation v2.0.0

This is a major update of the package, which most notably adds support for linear programming inflation problems. Major changes are:

• Added support for linear programming relaxations of causal scenarios, via InflationLP. This allows to run inflation hierarchies bounding the sets of classical and no-signaling correlations, per J. Causal Inference 7(2), 2019 (arXiv:1609.00672)
• Added support for hybrid scenarios with sources of different nature, via the classical_sources argument to InflationProblem. Currently supported: classical-no-signaling (via InflationLP) and classical-quantum (via InflationSDP).
• Added support for possibilistic-type feasibility problems (via supports_problem in InflationLP and InflationSDP).
• Added initial support for structures with multiple layers of latent variables.
• Improved support for structures with visible-to-visible connections, using the notation of do-conditionals.
• Improved handling of certificates. This makes them easier to manipulate and evaluate.
• Revamped the description of monomials. This makes the codes faster and consume less memory.

First release

This is the first release of the library.