coexist.AccessData

class coexist.AccessData(access_path='.')

Bases: object

Access (pun intended) data generated by a coexist.Access run; read it in using coexist.AccessData.read("access_seed<seed>").

Examples

Suppose you run coexist.Access.learn(random_seed = 123) - then a directory “access_seed123/” would be generated. Access (yes, still intended) all data generated in a Python-friendly format using:

>>> import coexist
>>> data = coexist.AccessData("access_123")
>>> data
AccessData
--------------------------------------------------------------------------
paths          ╎ AccessPaths(...)
parameters     ╎         value  min   max     sigma
               ╎ fp1 -0.005312 -5.0  10.0  0.024483
               ╎ fp2  0.003409 -5.0  10.0  0.034576
               ╎ fp3  0.296074 -5.0  10.0  2.078181
num_epochs     ╎ 31
target         ╎ 0.1
seed           ╎ 123
epochs         ╎ DataFrame(fp1_mean, fp2_mean, fp3_mean, fp1_std, fp2_std,
               ╎           fp3_std, overall_std)
epochs_scaled  ╎ DataFrame(fp1_mean, fp2_mean, fp3_mean, fp1_std, fp2_std,
               ╎           fp3_std, overall_std)
results        ╎ DataFrame(fp1, fp2, fp3, error)
results_scaled ╎ DataFrame(fp1, fp2, fp3, error)

Attributes:

pathsAccessPaths

Struct-like object storing relevant paths in the given ACCES directory.

parameterspd.DataFrame

The optimum free parameters found (final or intermediate if ACCES is still running).

parameters_scaledpd.DataFrame

The optimum free parameters found, divided by scaling such that the initial variance in the parameter values was unity.

scalingnp.ndarray

A vector with the values to scale each parameter by - they are the initial variances (sigma).

populationint

The number of simulations to run in parallel within a single epoch, or number of parameter combinations to try at once.

num_epochsint

The number of epochs that were successfully executed.

targetfloat

The target variance, where the initial parameter uncertainty must be decreased from 1 to target.

seedint

The random number generator seed uniquely defining this ACCES run.

epochsnp.ndarray

Matrix with columns [mean_param1, mean_param2, …, std_param1, std_param2, …, std_overall] with one row per epoch.

epochs_scalednp.ndarray

Same as epochs, scaled such that the initial variance (sigma) becomes unity.

resultsnp.ndarray

Matrix with columns [param1, param2, …, error] for each parameter combination tried - i.e. population * num_epochs.

results_scalednp.ndarray

Same as results, scaled such that the initial variance (sigma) becomes unity.

__init__(access_path='.')

Read in data generated by coexist.Access; the access_path can be either the “access_info_<hash>” directory itself, or its parent directory.

Methods

__init__([access_path])	Read in data generated by coexist.Access; the access_path can be either the "access_info_<hash>" directory itself, or its parent directory.
copy()	Return copy of AccessData object.
empty()	Create an empty AccessData object that you can set attributes to directly.
legacy(access_path)	Read in data from legacy coexist-0.1.0 ACCES format; this is normally called automatically by AccessData.read.
read([access_path])	Read in data generated by coexist.Access; the access_path can be either the "access_seed<hash>" directory itself, or its parent directory.
save(dirname)	Save AccessData to a new directory at dirname.

static empty()

Create an empty AccessData object that you can set attributes to directly.

Examples

Create an empty AccessData object:

>>> import coexist
>>> data = coexist.AccessData.empty()

static read(access_path='.')

Read in data generated by coexist.Access; the access_path can be either the “access_seed<hash>” directory itself, or its parent directory.

Here for backwards-compatibility; you can instantiate the class directly with the access_path, e.g. AccessData(".").

legacy(access_path)

Read in data from legacy coexist-0.1.0 ACCES format; this is normally called automatically by AccessData.read.

copy()

Return copy of AccessData object.

save(dirname)

Save AccessData to a new directory at dirname.