Reference#

Helpers#

`hopsy.compute_chebyshev_center`(problem[, ...])	Computes the Chebyshev center, that is the midpoint of a (non-unique) largest inscribed ball in the polytope defined by \(Ax \leq b\).
`hopsy.generate_unit_hypercube`(dimension)	Generate matrix A and vector b of the unit N-dimensional hypercube.
`hopsy.generate_unit_simplex`(dimension)	Generate matrix A and vector b of the unit N-dimensional simplex.
`hopsy.is_polytope_empty`(A, b[, S, h])	Checks whether the polytope given by Ax < b and optionally Sx=h has a solution x.
`hopsy.run_multiphase_sampling`(problem, ...)	runs multiphase sampling as suggested in https://drops.dagstuhl.de/opus/volltexte/2021/13820/pdf/LIPIcs-SoCG-2021-21.pdf limit_singular_value_ratio=2.3
`hopsy.simplify`(problem)	Simplifies the polytope defined in the `problem` by removing redundant constraints and refunction inequality constraints to equality constraints in case of dimension width less than thresh.

`hopsy.BirkhoffPolytope`(size)	Birkhoff polytope helper that manages transformation of states and constraints
`hopsy.LP`(args, *kwargs)	Singleton for controlling PolyRound parameters for Linear Programming.

`hopsy.Model`	Base model class.
`hopsy.Gaussian`(self[, mean, covariance, ...])	Gaussian model which can be invariant in some dimensions of the input vector.
`hopsy.Mixture`(self, components[, weights])	The `Mixture` is a weighted sum of \(n\) components, so its unnormalized density is given as
`hopsy.Rosenbrock`	A multi-dimensional Rosenbrock function in \(2n\) dimensions.

`hopsy.round`(problem[, simplify])	Rounds the polytope defined by the inequality \(Ax \leq b\) using PolyRound.
`hopsy.add_box_constraints`(problem, ...[, ...])	Adds box constraints to all dimensions.
`hopsy.add_equality_constraints`(problem, ...)	Adds equality constraints as specified.
`hopsy.back_transform`(problem, points)	Transforms samples back from the sampling space (typically rounded) to the original parameter space.
`hopsy.transform`(problem, points)	Transforms samples from the parameter space to the sampling space (typically rounded).

`hopsy.Proposal`	Interface for proposal mechanisms.
`hopsy.AdaptiveMetropolisProposal`	Adaptive Metropolis Algorithm for convex polytopes, see https://doi.org/10.2307/3318737 for unconstrained version.
`hopsy.BallWalkProposal`	This proposal is implemented for academic purpose and not recommended for production use, see https://doi.org/10.1007/s12532-015-0097-z for description.
`hopsy.BilliardWalkProposal`	See https://doi.org/10.3182/20140824-6-ZA-1003.02312 for descrption.
`hopsy.CSmMALAProposal`	Do not use when fisher information is expensive for your model.
`hopsy.DikinWalkProposal`	See https://doi.org/10.1016/j.orl.2016.07.005.
`hopsy.GaussianCoordinateHitAndRunProposal`	Mix of https://doi.org/10.1093/bioinformatics/btx052 and https://deepblue.lib.umich.edu/bitstream/handle/2027.42/3513/bal7884.0001.001.pdf.
`hopsy.GaussianHitAndRunProposal`	Mix of https://doi.org/10.1093/bioinformatics/btx052 and https://deepblue.lib.umich.edu/bitstream/handle/2027.42/3513/bal7884.0001.001.pdf.
`hopsy.GaussianProposal`	Random walk with gaussian steps.
`hopsy.ReversibleJumpProposal`	This is the implementation of the algorithm described in https://pubmed.ncbi.nlm.nih.gov/31214716/.
`hopsy.UniformCoordinateHitAndRunProposal`	See https://doi.org/10.1093/bioinformatics/btx052.
`hopsy.UniformHitAndRunProposal`	See: https://dl.acm.org/doi/pdf/10.1145/256562.256619
`hopsy.TruncatedGaussianProposal`	This is a gibbs sampler for constrained Gaussians, see https://doi.org/10.1093/bioinformatics/btz315.

`hopsy.RandomNumberGenerator`
`hopsy.Uniform`
`hopsy.Normal`

hopsy.MarkovChain(problem[, proposal, ...])

Given a hopsy.Problem a MarkovChain object can be constructed.

`hopsy.setup`(problem, random_seed[, ...])
`hopsy.sample`(markov_chains, rngs, n_samples)	Draw `n_samples` from every passed chain in `markov_chains` using the respective random number generator from `rngs`.

`hopsy.ess`(data[, series, method, relative, ...])	Calculate estimate of the effective sample size (ess).
`hopsy.mcse`(data[, series, method, prob, ...])	Calculate Markov Chain Standard Error statistic.
`hopsy.rhat`(data[, series, method, dask_kwargs])	Compute estimate of rank normalized splitR-hat for a set of traces.

hopsy.tune(mcs, rngs, target, n_tuning[, k, ...])

Thompson Sampling-based tuning method for specifying meaningful hyperparameters for Markov chain proposal distributions.