Profilers

minuit

class desilike.profilers.minuit.MinuitProfiler(*args, gradient=False, **kwargs)[source]

Bases: BaseProfiler

Wrapper for minuit profiler, used by the high-energy physics community for likelihood profiling.

Reference

Initialize profiler.

Parameters:

  • likelihood (BaseLikelihood) – Input likelihood.

  • rng (np.random.RandomState, default=None) – Random state. If None, seed is used to set random state.

  • seed (int, default=None) – Random seed.

  • max_tries (int, default=1000) – A ValueError is raised after this number of likelihood (+ prior) calls without finite posterior.

  • profiles (str, Path, Profiles) – Path to or profiles, to which new profiling results will be added.

  • ref_scale (float, default=1.) – Rescale parameters’ Parameter.ref reference distribution by this factor

  • rescale (bool, default=False) – If True, internally rescale parameters such their variation range is ~ unity. Provide covariance to take parameter variations from; else parameters’ Parameter.proposal will be used.

  • covariance (str, Path, ParameterCovariance, Chain, default=None) – If rescale, path to or covariance or chain, which is used for rescaling parameters. If None, parameters’ Parameter.proposal will be used instead.

  • gradient (bool, default=False) – If True, try to take the likelihood gradient (requires jax).

  • save_fn (str, Path, default=None) – If not None, save profiles to this location.

  • mpicomm (mpi.COMM_WORLD, default=None) – MPI communicator. If None, defaults to likelihood’s BaseLikelihood.mpicomm

contour(*args, **kwargs)[source]

Compute 2D contours for likelihood. The following attributes are added to profiles:

  • Profiles.contour

Parameters:

  • params (list, ParameterCollection, default=None) – List of tuples of parameters for which to compute 2D contours. If a list of parameters is provided instead, contours are computed for unique tuples of parameters.

  • cl (float, int, default=1) – Confidence level for the confidence contour. If not set or None, a standard 68.3 % confidence contour is produced. If 0 < cl < 1, the value is interpreted as the confidence level (a probability). If cl >= 1, it is interpreted as number of standard deviations. For example, cl = 3 produces a 3 sigma contour.

  • size (int, default=100) – Number of points on the contour to find. Increasing this makes the contour smoother, but requires more computation time.

  • interpolated (int, default=0) – Number of interpolated points on the contour. If you set this to a value larger than size, cubic spline interpolation is used to generate a smoother curve and the interpolated coordinates are returned. Values smaller than size are ignored. Good results can be obtained with size=20, interpolated=200.

covariance(**kwargs)

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.error # parabolic errors at best fit (if made available by the profiler)

  • Profiles.covariance # parameter covariance at best fit (if made available by the profiler).

grid(*args, **kwargs)[source]

Compute best fits on grid for likelihood. The following attributes are added to profiles:

  • Profiles.grid

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, dict, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, dict, default=1) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, dict, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

interval(*args, **kwargs)[source]

Compute confidence intervals for likelihood. The following attributes are added to profiles:

  • Profiles.interval

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to estimate confidence intervals.

  • cl (float, int, default=None) – Confidence level for the confidence interval. If not set or None, a standard 68.3 % confidence interval is produced. If 0 < cl < 1, the value is interpreted as the confidence level (a probability). If cl >= 1, it is interpreted as number of standard deviations. For example, cl = 3 produces a 3 sigma interval.

maximize(*args, **kwargs)[source]

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.start

  • Profiles.bestfit

  • Profiles.error # parabolic errors at best fit

  • Profiles.covariance # parameter covariance at best fit

One will typically run several independent likelihood maximizations in parallel, on number of MPI processes - 1 ranks (1 if single process), to make sure the global maximum is found.

Parameters:

  • niterations (int, default=None) – Number of iterations, i.e. of runs of the profiler from independent starting points. If None, defaults to mpicomm.size - 1 (if > 0, else 1).

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

profile(*args, **kwargs)[source]

Compute 1D profiles for likelihood. The following attributes are added to profiles:

  • Profiles.profile

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, default=30) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

bobyqa

class desilike.profilers.bobyqa.BOBYQAProfiler(likelihood, rng=None, seed=None, max_tries=1000, profiles=None, ref_scale=1.0, rescale=False, covariance=None, save_fn=None, mpicomm=None)[source]

Bases: BaseProfiler

Designed for solving bound-constrained general objective minimization, without requiring derivatives of the objective.

Reference

Initialize profiler.

Parameters:

  • likelihood (BaseLikelihood) – Input likelihood.

  • rng (np.random.RandomState, default=None) – Random state. If None, seed is used to set random state.

  • seed (int, default=None) – Random seed.

  • max_tries (int, default=1000) – A ValueError is raised after this number of likelihood (+ prior) calls without finite posterior.

  • profiles (str, Path, Profiles) – Path to or profiles, to which new profiling results will be added.

  • ref_scale (float, default=1.) – Rescale parameters’ Parameter.ref reference distribution by this factor

  • rescale (bool, default=False) – If True, internally rescale parameters such their variation range is ~ unity. Provide covariance to take parameter variations from; else parameters’ Parameter.proposal will be used.

  • covariance (str, Path, ParameterCovariance, Chain, default=None) – If rescale, path to or covariance or chain, which is used for rescaling parameters. If None, parameters’ Parameter.proposal will be used instead.

  • save_fn (str, Path, default=None) – If not None, save profiles to this location.

  • mpicomm (mpi.COMM_WORLD, default=None) – MPI communicator. If None, defaults to likelihood’s BaseLikelihood.mpicomm

contour(params=None, cl=1, niterations=1, **kwargs)

Compute 2D contours for likelihood. The following attributes are added to profiles:

  • Profiles.contour

Parameters:

  • params (list, ParameterCollection, default=None) – List of tuples of parameters for which to compute 2D contours. If a list of parameters is provided instead, contours are computed for unique tuples of parameters.

  • cl (float, int, default=1) – Confidence level for the confidence contour. If not set or None, a standard 68.3 % confidence contour is produced. If 0 < cl < 1, the value is interpreted as the confidence level (a probability). If cl >= 1, it is interpreted as number of standard deviations. For example, cl = 3 produces a 3 sigma contour.

  • **kwargs (dict) – Optional arguments for specific profiler.

covariance(**kwargs)

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.error # parabolic errors at best fit (if made available by the profiler)

  • Profiles.covariance # parameter covariance at best fit (if made available by the profiler).

grid(*args, **kwargs)[source]

Compute best fits on grid for likelihood. The following attributes are added to profiles:

  • Profiles.grid

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, dict, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, dict, default=1) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, dict, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

interval(params=None, cl=1, niterations=1, **kwargs)

Compute confidence intervals for likelihood. The following attributes are added to profiles:

  • Profiles.interval

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to estimate confidence intervals.

  • **kwargs (dict) – Optional arguments for specific profiler.

maximize(*args, **kwargs)[source]

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.start

  • Profiles.bestfit

  • Profiles.error # parabolic errors at best fit

  • Profiles.covariance # parameter covariance at best fit

One will typically run several independent likelihood maximizations in parallel, on number of MPI processes - 1 ranks (1 if single process), to make sure the global maximum is found.

Parameters:

  • niterations (int, default=None) – Number of iterations, i.e. of runs of the profiler from independent starting points. If None, defaults to mpicomm.size - 1 (if > 0, else 1).

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

  • npt (int, default=None) – The number of interpolation points to use; default is 2 * ndim + 1. Py-BOBYQA requires ndim + 1 <= npt <= (ndim + 1)(ndim + 2)/2. Larger values are particularly useful for noisy problems.

  • rhobeg (float, default=None) – The initial value of the trust region radius default is \(0.1 max(|x_0|_{\infty}, 1)\).

  • rhoend (float, default=1e-8) – Minimum allowed value of trust region radius, which determines when a successful termination occurs.

  • seek_global_minimum (bool, default=False) – A flag to indicate whether to search for a global minimum, rather than a local minimum. This is used to set some sensible default parameters, all of which can be overridden by the values provided in user_params. If True, both upper and lower bounds must be set. Note that Py-BOBYQA only implements a heuristic method, so there are no guarantees it will find a global minimum. However, by using this flag, it is more likely to escape local minima if there are better values nearby. The method used is a multiple restart mechanism, where we repeatedly re-initialize Py-BOBYQA from the best point found so far, but where we use a larger trust reigon radius each time (note: this is different to more common multi-start approach to global optimization).

profile(*args, **kwargs)[source]

Compute 1D profiles for likelihood. The following attributes are added to profiles:

  • Profiles.profile

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, default=30) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

scipy

class desilike.profilers.scipy.ScipyProfiler(*args, method=None, gradient=False, **kwargs)[source]

Bases: BaseProfiler

Wrapper for the collection of scipy’s profilers.

Reference

https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html#scipy.optimize.minimize

Initialize profiler.

Parameters:

  • likelihood (BaseLikelihood) – Input likelihood.

  • method (str, default=None) – Type of solver. Should be one of ‘Nelder-Mead’, ‘Powell’, ‘CG’, ‘BFGS’, ‘Newton-CG’, ‘L-BFGS-B’, ‘TNC’, ‘COBYLA’, ‘SLSQP’, ‘trust-constr’, ‘dogleg’, ‘trust-ncg’, ‘trust-exact’, ‘trust-krylov’. If not given, chosen to be one of ‘BFGS‘, ‘L-BFGS-B‘, ‘SLSQP‘, depending on whether or not sompe parameters have bounded priors.

  • gradient (bool, default=False) – Only for ‘CG‘, ‘BFGS‘, ‘Newton-CG‘, ‘L-BFGS-B‘, ‘TNC‘, ‘SLSQP‘, ‘dogleg‘, ‘trust-ncg‘, ‘trust-krylov‘, ‘trust-exact‘ and ‘trust-constr‘.

  • rng (np.random.RandomState, default=None) – Random state. If None, seed is used to set random state.

  • seed (int, default=None) – Random seed.

  • max_tries (int, default=1000) – A ValueError is raised after this number of likelihood (+ prior) calls without finite posterior.

  • profiles (str, Path, Profiles) – Path to or profiles, to which new profiling results will be added.

  • ref_scale (float, default=1.) – Rescale parameters’ Parameter.ref reference distribution by this factor

  • rescale (bool, default=False) – If True, internally rescale parameters such their variation range is ~ unity. Provide covariance to take parameter variations from; else parameters’ Parameter.proposal will be used.

  • covariance (str, Path, ParameterCovariance, Chain, default=None) – If rescale, path to or covariance or chain, which is used for rescaling parameters. If None, parameters’ Parameter.proposal will be used instead.

  • save_fn (str, Path, default=None) – If not None, save profiles to this location.

  • mpicomm (mpi.COMM_WORLD, default=None) – MPI communicator. If None, defaults to likelihood’s BaseLikelihood.mpicomm

contour(params=None, cl=1, niterations=1, **kwargs)

Compute 2D contours for likelihood. The following attributes are added to profiles:

  • Profiles.contour

Parameters:

  • params (list, ParameterCollection, default=None) – List of tuples of parameters for which to compute 2D contours. If a list of parameters is provided instead, contours are computed for unique tuples of parameters.

  • cl (float, int, default=1) – Confidence level for the confidence contour. If not set or None, a standard 68.3 % confidence contour is produced. If 0 < cl < 1, the value is interpreted as the confidence level (a probability). If cl >= 1, it is interpreted as number of standard deviations. For example, cl = 3 produces a 3 sigma contour.

  • **kwargs (dict) – Optional arguments for specific profiler.

covariance(**kwargs)

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.error # parabolic errors at best fit (if made available by the profiler)

  • Profiles.covariance # parameter covariance at best fit (if made available by the profiler).

grid(*args, **kwargs)[source]

Compute best fits on grid for likelihood. The following attributes are added to profiles:

  • Profiles.grid

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, dict, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, dict, default=1) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, dict, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

interval(params=None, cl=1, niterations=1, **kwargs)

Compute confidence intervals for likelihood. The following attributes are added to profiles:

  • Profiles.interval

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to estimate confidence intervals.

  • **kwargs (dict) – Optional arguments for specific profiler.

maximize(*args, **kwargs)[source]

Maximize likelihood. The following attributes are added to profiles:

  • Profiles.start

  • Profiles.bestfit

  • Profiles.error # parabolic errors at best fit (if made available by the solver)

  • Profiles.covariance # parameter covariance at best fit (if made available by the solver).

One will typically run several independent likelihood maximizations in parallel, on number of MPI processes - 1 ranks (1 if single process), to make sure the global maximum is found.

Parameters:

  • niterations (int, default=None) – Number of iterations, i.e. of runs of the profiler from independent starting points. If None, defaults to mpicomm.size - 1 (if > 0, else 1).

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.

  • tol (float, default=None) – Tolerance for termination. When tol is specified, the selected minimization algorithm sets some relevant solver-specific tolerance(s) equal to tol. For detailed control, use solver-specific options.

  • kwargs (dict) – Solver-specific options.

profile(*args, **kwargs)[source]

Compute 1D profiles for likelihood. The following attributes are added to profiles:

  • Profiles.profile

Parameters:

  • params (str, Parameter, list, ParameterCollection, default=None) – Parameters for which to compute 1D profiles.

  • grid (array, list, default=None) – Parameter values on which to compute the profile, for each parameter. If grid is set, size and bound are ignored.

  • size (int, list, default=30) – Number of scanning points. Ignored if grid is set. Can be specified for each parameter.

  • cl (int, list, default=2) – If bound is a number, it specifies an interval of N sigmas symmetrically around the minimum. Ignored if grid is set. Can be specified for each parameter.

  • niterations (int, default=1) – Number of iterations, i.e. of runs of the profiler from independent starting points.

  • max_iterations (int, default=int(1e5)) – Maximum number of likelihood evaluations.