Source code for pyretis.inout.analysisio.analysisio

#! /usr/bin/env python3
# -*- coding: utf-8 -*-
# Copyright (c) 2023, PyRETIS Development Team.
# Distributed under the LGPLv2.1+ License. See LICENSE for more info.
"""Methods that will output results from the analysis functions.

The Methods defined here will also run the analysis and output
according to given settings.

Important methods defined here
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

analyse_file (:py:func:`.analyse_file`)
    Method to analyse a file. For example, it can be used as

    >>> from pyretis.inout.analysisio import analyse_file
    >>> result, raw_data = analyse_file('cross', 'cross.txt', settings)

    To output the results, for instance by plotting, one can do:

    >>> figures = output_results(file_type, plotter, result, raw_data)
    >>> txt_file = output_results(file_type, txt_plotter, result, raw_data)


run_analysis_files (:py:func:`.run_analysis_files`)
    Methods to the analysis on a set of files. It will create output
    that can be used for reporting.
"""
import logging
import os
import numpy as np
from pyretis.core.units import create_conversion_factors
from pyretis.core.pathensemble import generate_ensemble_name
from pyretis.analysis import (analyse_flux, analyse_energies, analyse_orderp,
                              analyse_path_ensemble, analyse_repptis_ensemble,
                              match_probabilities, retis_flux, retis_rate,
                              perm_calculations)
from pyretis.inout import print_to_screen
from pyretis.inout.formats.formatter import format_number
from pyretis.inout.plotting import create_plotter
from pyretis.inout.plotting import TxtPlotter
from pyretis.inout.report import generate_report
from pyretis.inout.settings import SECTIONS, copy_settings
from pyretis.inout.formats.pathensemble import PathEnsembleFile
from pyretis.inout.formats.energy import EnergyFile
from pyretis.inout.formats.order import OrderFile
from pyretis.inout.formats.cross import CrossFile
from pyretis.inout.simulationio import OUTPUT_TASKS
logger = logging.getLogger(__name__)  # pylint: disable=invalid-name
logger.addHandler(logging.NullHandler())

np.set_printoptions(legacy='1.25')

__all__ = ['analyse_file', 'run_analysis_files',
           'run_analysis', 'read_first_block',
           'repptis_running_pcross_analysis']


_ANALYSIS_FUNCTIONS = {'cross': analyse_flux,
                       'order': analyse_orderp,
                       'energy': analyse_energies,
                       'pathensemble': analyse_path_ensemble,
                       'pathensemble_repptis': analyse_repptis_ensemble}


# Input files for analysis
_FILES = {'md-flux': {},
          'md-nve': {},
          'md': {},
          'tis': {},
          'make-tis-files': {},
          'retis': {},
          'repptis': {}}
# Add files
for key in ('cross', 'energy', 'order'):
    _FILES['md-flux'][key] = OUTPUT_TASKS[key]['filename']
for key in ('energy',):
    _FILES['md-nve'][key] = OUTPUT_TASKS[key]['filename']
    _FILES['md'][key] = OUTPUT_TASKS[key]['filename']
for key in ('pathensemble',):
    for key2 in ('make-tis-files', 'tis', 'retis', 'repptis'):
        _FILES[key2][key] = OUTPUT_TASKS[key]['filename']


[docs]def run_analysis(settings): """Run a predefined analysis task. Parameters ---------- settings : dict Simulation settings and settings for the analysis. Returns ------- out : dict A dictionary with the results from the analysis. This dict can be used to generate a report. """ runners = {'retis': run_retis_analysis, 'tis': run_tis_analysis, 'make-tis-files': run_tis_analysis, 'md-flux': run_mdflux_analysis, 'repptis': run_repptis_analysis} sim_task = settings['simulation']['task'].lower() report_dir = settings['analysis'].get('report-dir', None) plotter = create_plotter(settings['analysis']['plot'], out_dir=report_dir) backup = settings['output'].get('backup', SECTIONS['output']['backup']) txt_plotter = TxtPlotter(settings['analysis']['txt-output'], backup=backup, out_dir=report_dir) if sim_task in runners.keys(): runner = runners[sim_task] return runner(settings, plotter, txt_plotter) msgtxt = f'Unknown analysis task "{sim_task}" requested!' logger.error(msgtxt) raise ValueError(msgtxt)
def get_path_ensemble_files(ensemble_number, settings, detect, interfaces): """Return files for a single path ensemble. Here, we will return the files needed to analyse a single path ensemble and we will also return settings which can be used for the analysis. Parameters ---------- ensemble_number : int This is the integer representing the ensemble. settings : dict The settings to use for an analysis/simulation detect : float or None The interface use for detecting if a path is successful for not. interfaces : list of floats The interfaces used for this particular path simulation. Returns ------- local_settings : dict This dict contains settings which can be used for a initial flux analysis. files : list of tuples The tuples in this list are the files which can be analysed further, using the settings in `out[0]`. """ sim_task = settings['simulation']['task'].lower() lsetting = copy_settings(settings) lsetting['simulation']['interfaces'] = interfaces lsetting['tis']['detect'] = detect lsetting['tis']['ensemble_number'] = ensemble_number run_path = settings['simulation']['exe-path'] files = [] for file_type in _FILES[sim_task]: filename = os.path.join(run_path, generate_ensemble_name(ensemble_number), _FILES[sim_task][file_type]) if os.path.isfile(filename): files.append((file_type, filename)) logger.debug('Adding file "%s" for analysis', filename) else: logger.debug('Could not find file %s', filename) return lsetting, files def get_path_simulation_files(sim_settings): """Set up for analysis of TIS and RETIS simulations. For these kinds of simulations, we expect to analyse several directories with raw-data. For TIS we expect to find a directory with raw-data for the initial flux (named ``flux``) and the directories for the path simulations (named ``001``, ``002`` etc. for ensembles [0^+], [1^+] and so on). For RETIS, we expect to find the same directories but with a ``000`` (for the ``[0^-]`` ensemble) rather than the ``flux`` directory. Parameters ---------- sim_settings : dict The settings used for the (RE)TIS simulation(s). These settings are used to get the interfaces used and the path ensembles defined by these interfaces. Returns ------- all_settings : list of dict This dict in this list contains settings which can be used for the analysis. all_files : list of lists of tuples `all_files[i]` is a list of files from path ensemble simulation `i`. For TIS, `all_files[0]` should be the files obtained in the initial flux simulation. These files can be analysed using the settings in `all_settings[i]`. """ # Check if we can do flux analysis: all_files, all_settings = [], [] interfaces = sim_settings['simulation']['interfaces'] reactant = interfaces[0] product = interfaces[-1] task = sim_settings['simulation']['task'] # for PPTIS or REPPTIS: memory mem = sim_settings.get(task, {}).get('memory', 1) if task == 'make-tis-files': all_files.append(None) all_settings.append(None) else: # just add the 0 ensemble detect = None interface = [-float('inf'), reactant, reactant] setts, files = get_path_ensemble_files(0, sim_settings, detect, interface) all_files.append(files) all_settings.append(setts) for i, middle in enumerate(interfaces[:-1]): detect = interfaces[i + 1] if i < len(interfaces) else product if task in ['pptis', 'repptis']: reactant = interfaces[max(0, i - mem)] product = interfaces[min(i + mem, len(interfaces) - 1)] interface = [reactant, middle, product] else: interface = [reactant, middle, product] setts, files = get_path_ensemble_files(i + 1, sim_settings, detect, interface) all_files.append(files) all_settings.append(setts) return all_settings, all_files def print_value_error(heading, value, rel_error, level=None): """Print out the matched probabilities.""" val = format_number(value, 0.1, 100) msgtxt = f'{heading}: {val}' print_to_screen(msgtxt.strip(), level=level) fmt_scale = format_number(rel_error * 100, 0.1, 100) msgtxt = f'(Relative error: {fmt_scale.rstrip()} %)' print_to_screen(msgtxt, level=level) def run_single_tis_analysis(settings, plotter, txt_plotter): """Run the analysis for a single TIS ensemble. Parameters ---------- settings : dict The settings to use for an analysis/simulation. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. Returns ------- out : list or dict or similar The output from the analysis. """ sim = settings['simulation'] tis = settings['ensemble'][0]['tis'] sett, files = get_path_ensemble_files( tis['ensemble_number'], settings, tis.get('detect', sim['interfaces'][-1]), sim['interfaces']) msgtxt = f"Analysing ensemble {tis['ensemble_number']}" print_to_screen(msgtxt, level='info') print_to_screen() result = run_analysis_files(sett, files, plotter, txt_plotter) report_txt = generate_report('tis', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() return result def run_tis_analysis(settings, plotter, txt_plotter): """Run the analysis for TIS. Parameters ---------- settings : dict The settings to use for an analysis/simulation. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. Returns ------- out : list or dict or similar The output from the analysis. """ list_ens = settings['ensemble'] if len(list_ens) == 1 and settings['simulation']['task'] == 'tis': return run_single_tis_analysis(settings, plotter, txt_plotter) results = {'cross': None, 'pathensemble': [], 'matched': None} all_settings, all_files = get_path_simulation_files(settings) nens = len(all_settings) - 1 for i, (sett, files) in enumerate(zip(all_settings, all_files)): if i == 0: msgtxt = ('Initial flux is not calculated here.\n' 'Remember to calculate this separately!') logger.info(msgtxt) print_to_screen(msgtxt, level='warning') else: msgtxt = f'Analysing ensemble {i} of {nens}' print_to_screen(msgtxt, level='info') print_to_screen() result = run_analysis_files(sett, files, plotter, txt_plotter) results['pathensemble'].append(result['pathensemble']) report_txt = generate_report('tis', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() # match probabilities: out, fig, txt = analyse_and_output_matched(results['pathensemble'], plotter, txt_plotter, settings) results['matched'] = {'out': out, 'figures': fig, 'txtfile': txt} print_to_screen('Overall results', level='success') print_to_screen('===============', level='success') print_to_screen() print_value_error('TIS Crossing probability', out['prob'], out['relerror'], level='success') return results def run_retis_analysis(settings, plotter, txt_plotter): """Run the analysis for RETIS. Parameters ---------- settings : dict The settings to use for an analysis/simulation. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. """ units = settings['system']['units'] permeability = settings['simulation'].get('permeability', False) create_conversion_factors(units, **settings.get('unit-system', {})) all_settings, all_files = get_path_simulation_files(settings) results = {'cross': None, 'pathensemble': [], 'matched': None, 'permeability': permeability} nens = len(all_settings) - 1 print_to_screen() for i, (sett, files) in enumerate(zip(all_settings, all_files)): msgtxt = f'Analysing ensemble {i} of {nens}' print_to_screen(msgtxt, level='info') print_to_screen() if i == 0: result = run_analysis_files(sett, files, plotter, txt_plotter) results['pathensemble0'] = result['pathensemble'] report_txt = generate_report('retis0', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() else: result = run_analysis_files(sett, files, plotter, txt_plotter) results['pathensemble'].append(result['pathensemble']) report_txt = generate_report('tis', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() # flux first: time_subcycles = settings['engine'].get('subcycles', 1) timestep = settings['engine']['timestep'] * time_subcycles flux, flux_error = retis_flux(results['pathensemble0']['out'], results['pathensemble'][0]['out'], timestep) results['flux'] = {'value': flux, 'error': flux_error, 'unit': units} # match probabilities: out, fig, txt = analyse_and_output_matched(results['pathensemble'], plotter, txt_plotter, settings, flux=flux) results['matched'] = {'out': out, 'figures': fig, 'txtfile': txt} rate, rate_error = retis_rate(out['prob'], out['relerror'], flux, flux_error) results['rate'] = {'value': rate, 'error': rate_error, 'unit': units} xi, tau, perm, xi_err, tau_err, perm_err = perm_calculations( results['pathensemble0'], out['prob'], out['relerror']) xi_figs, tau_figs = perm_figures(results['pathensemble0'], plotter) results['xi'] = {'value': xi, 'error': xi_err, 'fig': xi_figs} results['tau'] = {'value': tau, 'error': tau_err, 'fig': tau_figs} results['perm'] = {'value': perm, 'error': perm_err} print_to_screen('Overall results', level='success') print_to_screen('===============', level='success') print_to_screen() print_value_error('RETIS Crossing probability', out['prob'], out['relerror'], level='success') print_to_screen() print_value_error(f'Initial flux (units 1/{units})', flux, flux_error, level='success') print_to_screen() print_value_error(f'Rate constant (units 1/{units})', rate, rate_error, level='success') if permeability: print_to_screen() print_value_error('Xi', xi, xi_err, level='success') print_to_screen() print_value_error(f'Tau/dz (unit: {units}/OP-unit)', tau, tau_err, level='success') print_to_screen() print_value_error(f'Permeability (unit: OP-unit/{units})', perm, perm_err, level='success') return results def run_repptis_analysis(settings, plotter, txt_plotter): """Run the analysis for PPRETIS. Parameters ---------- settings : dict The settings to use for an analysis/simulation. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. """ units = settings['system']['units'] permeability = settings['simulation'].get('permeability', False) create_conversion_factors(units, **settings.get('unit-system', {})) all_settings, all_files = get_path_simulation_files(settings) results = {'cross': None, 'pathensemble': [], 'matched': None, 'permeability': permeability, 'reptis': {}} nens = len(all_settings) - 1 print_to_screen() probs = [] ptypes = [] for i, (sett, files) in enumerate(zip(all_settings, all_files)): msgtxt = f'Analysing ensemble {i} of {nens}' # We change filetype to repptis instead of retis for anaylsis repptis_str = 'pathensemble_repptis' files0 = [(repptis_str, files[0][1])] print_to_screen(msgtxt, level='info') print_to_screen() if i == 0: result = run_analysis_files(sett, files0, plotter, txt_plotter) results['pathensemble0'] = result['pathensemble_repptis'] report_txt = generate_report('retis0', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() else: result = run_analysis_files(sett, files0, plotter, txt_plotter) results['pathensemble'].append(result['pathensemble_repptis']) pp_lmr = result[repptis_str]['out']['prun_sl'][-1] pp_rml = result[repptis_str]['out']['prun_sr'][-1] ptype = result[repptis_str]['out']['ptypes'] ptypes.append(ptype) pp_lml = 1 - pp_lmr pp_rmr = 1 - pp_rml pp_lm_err = result[repptis_str]['out']['blockerror_sl'][2] pp_rm_err = result[repptis_str]['out']['blockerror_sr'][2] probs.append({"LMR": {"mean": pp_lmr, "std": pp_lm_err}, "RML": {"mean": pp_rml, "std": pp_rm_err}, "RMR": {"mean": pp_rmr, "std": pp_rm_err}, "LML": {"mean": pp_lml, "std": pp_lm_err}, }) report_txt = generate_report('pptis', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() # flux first: time_subcycles = settings['engine'].get('subcycles', 1) timestep = settings['engine']['timestep'] * time_subcycles flux, flux_error = retis_flux(results['pathensemble0']['out'], results['pathensemble'][0]['out'], timestep) results['flux'] = {'value': flux, 'error': flux_error, 'unit': units} results.update(repptis_running_pcross_analysis(ptypes)) figures = plotter.output_matched_probability( [], settings['simulation']['interfaces'], results, reptis=True ) txt_plotter.output_matched_probability( [], settings['simulation']['interfaces'], results, reptis=True ) results['matched_fig'] = figures results['cross'] = {'value': results['pcrossrun'][-1], 'error': results['pcrossrun_hav_rele']} rate, rate_error = retis_rate(results['pcrossrun'][-1], results['pcrossrun_hav_rele'], flux, flux_error) results['rate'] = {'value': rate, 'error': rate_error, 'unit': units} xi, tau, perm, xi_err, tau_err, perm_err = perm_calculations( results['pathensemble0'], results['pcrossrun'][-1], results['pcrossrun_hav_rele']) xi_figs, tau_figs = perm_figures(results['pathensemble0'], plotter) results['xi'] = {'value': xi, 'error': xi_err, 'fig': xi_figs} results['tau'] = {'value': tau, 'error': tau_err, 'fig': tau_figs} results['perm'] = {'value': perm, 'error': perm_err} print_to_screen('Overall results', level='success') print_to_screen('===============', level='success') print_to_screen() print_value_error('REPPTIS Crossing probability', results['pcrossrun'][-1], results['pcrossrun_hav_rele'], level='success') print_to_screen() print_value_error(f'Initial flux (units 1/{units})', flux, flux_error, level='success') print_to_screen() print_value_error(f'Rate constant (units 1/{units})', rate, rate_error, level='success') if permeability: print_to_screen() print_value_error('Xi', xi, xi_err, level='success') print_to_screen() print_value_error(f'Tau/dz (unit: {units}/OP-unit)', tau, tau_err, level='success') print_to_screen() print_value_error(f'Permeability (unit: OP-unit/{units})', perm, perm_err, level='success') return results def run_mdflux_analysis(settings, plotter, txt_plotter): """Run the analysis for the md flux simulation. Parameters ---------- settings : dict The settings to use for an analysis/simulation. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. Returns ------- out : list or dict or similar The output from the analysis. """ sim = settings['simulation'] sim_task = sim['task'] exe_path = settings['simulation'].get('exe-path', os.getcwd()) files = [] for file_type in _FILES[sim_task]: filename = _FILES[sim_task][file_type] if os.path.isfile(os.path.join(exe_path, filename)): files.append((file_type, os.path.join(exe_path, filename))) msgtxt = 'Running analysis of a MD flux simulation...' print_to_screen(msgtxt, level='info') print_to_screen() result = run_analysis_files(settings, files, plotter, txt_plotter) report_txt = generate_report('md-flux', result, output='txt')[0] print_to_screen(''.join(report_txt)) print_to_screen() return result
[docs]def run_analysis_files(settings, files, plotter, txt_plotter): """Run the analysis on a collection of files. Parameters ---------- settings : dict This dict contains settings which dictate how the analysis should be performed and it should also contain information on how the simulation was performed. files : list of tuples This list contains the raw files to be analysed. The tuples are on format ('filetype', 'filename'). plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. Returns ------- out: dict The results from the analysis. """ results = {} for (file_type, file_name) in files: result, raw_data = analyse_file(file_type, file_name, settings) figures = output_results(file_type, plotter, result, raw_data) txt_file = output_results(file_type, txt_plotter, result, raw_data) results[file_type] = {'out': result, 'figures': figures, 'txtfile': txt_file} return results
[docs]def read_first_block(file_type, file_name): """Read the first block of data from a file. This method will read the first block of data from a file and immediately return. Parameters ---------- file_type : string A string used for selecting the file object to use here. file_name : string The file to open. Returns ------- out : numpy.array The raw data read from the file. """ first_block = None class_map = {'energy': EnergyFile, 'order': OrderFile, 'cross': CrossFile} if file_type not in class_map: raise ValueError('Unknown file type requested in analysis.') klass = class_map[file_type] with klass(file_name, 'r') as fileobj: for block in fileobj.load(): if first_block is None: first_block = block else: logger.warning( ( 'Noticed a second block in the input file "%s".\n' 'This will be ignored by the flux analysis.\n' 'Are you are running the analysis with ' 'the correct input?' ), file_name ) break if first_block is None: return None return first_block['data']
def output_results(file_type, plotter, result, rawdata): """Plot the results. Plotting might here refer to both figures and text. Parameters ---------- file_type : string This determines what we are going to plot. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. result : list of lists or dicts This contains the results from a specific analysis. rawdata : list of floats, lists or objects The raw data with analysis results. Returns ------- out : list of strings These are the files that the plotter created. """ if file_type == 'cross': return plotter.output_flux(result) if file_type == 'order': return plotter.output_orderp(result, rawdata) if file_type == 'energy': return plotter.output_energy(result, rawdata) if file_type == 'pathensemble': return plotter.output_path(result, rawdata) if file_type == 'pathensemble_repptis': return plotter.output_pppath(result, rawdata)
[docs]def analyse_file(file_type, file_name, settings): """Run analysis on the given file. This function is included for convenience so that we can call an analysis like ``analyse_file('cross', 'cross.txt')`` i.e. it should automatically open the file and apply the correct analysis according to a given file type. Here we return a function to do the analysis, so we are basically wrapping one of the analysis functions. This is done in case we wish to rerun the analysis but with different settings for instance. Parameters ---------- file_type : string This is the type of file we are to analyse. file_name : string The file name to open. settings : dict This dict contains settings which dictate how the analysis should be performed and information on how the simulation was performed. Returns ------- results : list or dict The output from the analysis raw_data : list, numpy.array or another type of object The raw data used in the analysis. """ function = _ANALYSIS_FUNCTIONS.get(file_type, None) if function is None: msgtxt = f'Unknown file type "{file_type}" requested!' logger.error(msgtxt) raise ValueError(msgtxt) results, raw_data = None, None if file_type in ('energy', 'order', 'cross'): # Read the first block of raw data and analyse it: raw_data = read_first_block(file_type, file_name) results = function(raw_data, settings) elif file_type in ('pathensemble', 'pathensemble_repptis'): # Create a PathEnsembleFile object to use for the analysis. interfaces = settings['simulation']['interfaces'] ensemble_settings = { 'ensemble_number': settings['tis']['ensemble_number'], 'interfaces': interfaces, 'detect': settings['tis'].get('detect', interfaces[1]) } raw_data = PathEnsembleFile(file_name, 'r', ensemble_settings) results = function(raw_data, settings) return results, raw_data
def analyse_and_output_matched(raw_data, plotter, txt_plotter, settings, flux=None): """Analyse and output matched probability. This will calculate the over-all crossing probability by combining results from many path simulations. Parameters ---------- raw_data : list, numpy.array or another type of object The raw data used in the analysis. plotter : object like :py:class:`.Plotter` This is the object that handles the plotting. txt_plotter : object like :py:class:`.Plotter` This is the object that handles the text output. settings : dict This dictionary contains settings for the analysis. Here we make use of the following keys from the analysis section: * `ngrid`: The number of grid points for calculating the crossing probability as a function of the order parameter. * `maxblock`: The max length of the blocks for the block error analysis. Note that this will maximum be equal the half of the length of the data, see `block_error` in `.analysis`. * `blockskip`: Can be used to skip certain block lengths. A `blockskip` equal to `n` will consider every n'th block up to `maxblock`, i.e. it will use block lengths equal to `1`, `1+n`, `1+2n`, etc. * `bins`: The number of bins to use for creating histograms. flux : float, optional The computed flux to be used only in RETIS if the running average of the computed rate is desired. Returns ------- out[0] : dict This dict contains the results from the analysis out[1] : list of dicts A dictionary with the figure files created (if any). out[2] : list of strings A list with the text files created (if any). """ path_results, ensemble_names, detect = [], [], [] interface_left = None for ensemble in raw_data: path_results.append(ensemble['out']) ensemble_names.append(ensemble['out']['ensemble']) detect.append(ensemble['out']['detect']) if interface_left is None: interface_left = ensemble['out']['interfaces'][0] result = match_probabilities(path_results, detect, settings) if flux is not None: result['overall-rrun'] = flux*result['overall-prun'] # for the figure, we add the A interface: detect_plot = [interface_left] + detect figures = plotter.output_matched_probability(ensemble_names, detect_plot, result) outtxt = txt_plotter.output_matched_probability(ensemble_names, detect, result) return result, figures, outtxt def perm_figures(results0, plotter): out = results0['out'] # Quick check to see if we need to do something if 'xi' not in out: return None, None xi = out['xi'] xierror = out['xierror'] bin_edges = out['tau_bin'] tau = out['tau'] tauerror = out['tauerror'] tau_ref = out['tau_ref'] tau_ref_bins = out['tau_ref_bins'] xi_figures = plotter.output_xi(xi, xierror) tau_figures = plotter.output_tau(tau, tauerror, bin_edges, tau_ref, tau_ref_bins) return xi_figures, tau_figures
[docs]def repptis_running_pcross_analysis(l_ptypes): """Recursive block analysis for repptis. Parameters ---------- l_ptypes : list List of cumsum ptypes for the ensembles [0+-'], [1+-], ... [N-1+-]. Returns ------- results : dict The results from the analysis. """ l_locps = [] for ptypes in l_ptypes: # loop over ensembles # if a division by zero occurs, we set the probability to NaN # casting to floats for true_divide ptypes = ptypes.astype(float) pmps = np.divide(ptypes[:, 0], ptypes[:, 0]+ptypes[:, 1], out=np.full_like(ptypes[:, 0], np.nan), where=(ptypes[:, 0]+ptypes[:, 1]) != 0) pmms = np.divide(ptypes[:, 1], ptypes[:, 0]+ptypes[:, 1], out=np.full_like(ptypes[:, 0], np.nan), where=(ptypes[:, 0]+ptypes[:, 1]) != 0) ppps = np.divide(ptypes[:, 2], ptypes[:, 2]+ptypes[:, 3], out=np.full_like(ptypes[:, 2], np.nan), where=(ptypes[:, 2]+ptypes[:, 3]) != 0) ppms = np.divide(ptypes[:, 3], ptypes[:, 2]+ptypes[:, 3], out=np.full_like(ptypes[:, 2], np.nan), where=(ptypes[:, 2]+ptypes[:, 3]) != 0) # we have to_list them, as we will append them until all have N_max cyc l_locps.append([pmms.tolist(), pmps.tolist(), ppps.tolist(), ppms.tolist()]) # So, l_locps: list of lists of lists: n_ensembles x 4 x n_blocks # extend the l_locps to the same length. We copy the last element of the # list until the length is the same as the longest list max_len = max([len(ll[0]) for ll in l_locps]) for ll in l_locps: while len(ll[0]) < max_len: for i in range(4): ll[i].append(ll[i][-1]) # We can now calculate the global crossing probabilities, for each cycle pcross_run = [] for icyc in range(max_len): # We have to get the local crossing probabilities for each ensemble l_pcross = [[ll[i][icyc] for i in range(4)] for ll in l_locps] # so l_pcross is shape n_ensembles x 4, but we need 4 lists of # length n_ensembles l_pcross = list(map(list, zip(*l_pcross))) # Now we calculate the global crossing probabilities _, _, pcross = get_global_probz(*l_pcross) # We append the last element of pcross to the running list pcross_run.append(pcross[-1] if pcross is not np.nan else np.nan) # Now do the recursive block analysis rel_errors, nstatineff, half_av_err = recursive_block_analysis(pcross_run) results = {'pcross': pcross, 'pcrossrun': pcross_run, 'pcrossrun_rele': rel_errors, 'nstatineff': nstatineff, 'pcrossrun_hav_rele': half_av_err} # make into retis format results['overall-cycle'] = list(range(len(pcross_run))) results['overall-prun'] = pcross_run results['overall-error'] = [list(range(len(rel_errors))), 0, 0, rel_errors, rel_errors[-1]] return results
def get_global_probz(pmps, pmms, ppps, ppms): """Return the global crossing probabilities for the PPTIS simulation. This follows the recursive relation P_plus[j] = (pLMR[j-1]*P_plus[j-1]) / (pLMR[j-1]+pLML[j-1]*P_min[j-1]) P_min[j] = (pRML[j-1]P_min[j-1])/(pLMR[j-1]+pLML[j-1]*P_min[j-1]) where P_plus[0] = 1, P_min[0] = 1, and j is the ensemble number. Parameters ---------- pmps : list of floats The local probability (P) of having type LMR (MinusPlus) for each ensemble (S). pmms : list of floats Local LML ppps : list of floats Local RMR ppms : list of floats Local RML Returns ------- pmin : list of floats Float per ensemble i. Represents the probability of crossing A earlier than i+1, given that the path crossed i. pplus : list of floats Float per ensemble i. Represents the probability of crossing i+1 earlier than A, given that the path crossed i. pcross : list of floats Float per ensemble i. Represents the TIS probability of crossing i+1. """ # if there is any NaN in pmps, pmms, ppps, ppms, return NaN if np.isnan(pmps).any() or np.isnan(pmms).any() or \ np.isnan(ppps).any() or np.isnan(ppms).any(): return [np.nan, np.nan, np.nan] pplus, pmin, pcross = [1.], [1.], [1., pmps[0]] for i, pmp, pmm, _, ppm in zip(range(len(pmps)), pmps, pmms, ppps, ppms): if i == 0: # This is [0^{\pm}'], so skip continue # Calculate the global crossing probabilities # if divide by zero, or divide by NaN, return NaN if pmp + pmm * pmin[-1] == 0 or\ (pmp is np.nan) or\ (pmm * pmin[-1] is np.nan): return [np.nan, np.nan, np.nan] pplus.append((pmp*pplus[-1])/(pmp+pmm*pmin[-1])) pmin.append((ppm*pmin[-1])/(pmp+pmm*pmin[-1])) # Calculate the TIS probabilities pcross.append(pmps[0]*pplus[-1]) return pmin, pplus, pcross def recursive_block_analysis(flist, minblocks=5): """Run a recursive block analysis on (time-series) data f. Parameters ---------- flist : list of floats The data to analyse. minblocks : int, optional The minimum number of blocks to use for the analysis. Returns ------- rel_errors : list of floats The relative errors for each block length. nstatineff : float The statistical inefficiency. half_av_err : float The average error for the second half of the relative errors. """ # flist is a list of floats, but we need to remove the np.nan's and 0's flist = np.array(flist) flist = flist[~np.isnan(flist)] flist = flist[flist != 0] # We now run the recursive blocking analysis # If we don't have enough data, we don't want the entire analysis to # fail. So just return NaNs, such that we still get those beautiful local # plots in the report PDF. if len(flist) < minblocks: return [np.nan], np.nan, np.nan bestavg = flist[-1] maxblocklen = int(len(flist)/minblocks) rel_errors = [] for blocklen in range(1, maxblocklen+1): runav_red = flist[blocklen-1::blocklen] blocks = recursive_blocks(runav_red) sum_quad_diff = np.sum(np.fromiter(((b-bestavg)**2 for b in blocks), dtype=float)) nb = len(blocks) aerr2 = sum_quad_diff/(nb*(nb-1)) aerr = np.sqrt(aerr2) rerr = aerr/bestavg rel_errors.append(rerr) # We will now take the mean of the second half of the relative errors half_av_err = np.mean(rel_errors[int(len(rel_errors)/2):]) nstatineff = (half_av_err/rel_errors[0])**2 # Now we have to add the NaN's back in, as 0's return rel_errors, nstatineff, half_av_err def recursive_blocks(recu): blocks = [] for i in range(len(recu)): if i == 0: blocks.append(recu[i]) else: blocks.append((i + 1) * recu[i] - i * recu[i - 1]) return blocks