PEMtk fitting setup & batch run demo

06/06/21

Outline of this notebook:

Use setup_fit_demo.py script to load data and setup fitting environment.
Run a batch of fits…
- 1. run batch
- 1. load a batch
Exploring the results… see the analysis notebook for more.

Setup

[1]:

# Load demo fitting workspace

# Run defaults
# %run 'setup_fit_demo.py'
# %run '/home/femtolab/github/PEMtk/demos/fitting/setup_fit_demo.py'

# With package path root passed
# %run '/home/femtolab/github/PEMtk/demos/fitting/setup_fit_demo.py' '<path-to-packages>'

# Quick hack for Stimpy
%run "D:\code\github\PEMtk\demos\fitting\setup_fit_demo.py"

*** Setting up demo fitting workspace and main `data` class object...
For more details see https://pemtk.readthedocs.io/en/latest/fitting/PEMtk_fitting_basic_demo_030621-full.html


* Loading packages...

* Importing local packages from root D:\code\github. Pass search path to the script if this fails.


* Loading demo matrix element data from D:\code\github\ePSproc\data\photoionization\n2_multiorb...

*** Job orb6 details
Key: orb6
Dir D:\code\github\ePSproc\data\photoionization\n2_multiorb, 1 file(s).
{   'batch': 'ePS n2, batch n2_1pu_0.1-50.1eV, orbital A2',
    'event': ' N2 A-state (1piu-1)',
    'orbE': -17.096913836366,
    'orbLabel': '1piu-1'}

*** Job orb5 details
Key: orb5
Dir D:\code\github\ePSproc\data\photoionization\n2_multiorb, 1 file(s).
{   'batch': 'ePS n2, batch n2_3sg_0.1-50.1eV, orbital A2',
    'event': ' N2 X-state (3sg-1)',
    'orbE': -17.341816310545997,
    'orbLabel': '3sg-1'}


* Loading demo ADM data from D:\code\github\ePSproc\data\alignment\N2_ADM_VM_290816.mat...

* Subselecting data...
Subselected from dataset 'orb5', dataType 'matE': 36 from 11016 points (0.33%)
Subselected from dataset 'pol', dataType 'pol': 1 from 3 points (33.33%)
Subselected from dataset 'ADM', dataType 'ADM': 52 from 14764 points (0.35%)

* Calculating MF-BLMs...
Subselected from dataset 'sim', dataType 'AFBLM': 195 from 195 points (100.00%)

*Setting  up fit parameters (with constraints)...
Set 6 complex matrix elements to 12 fitting params, see self.params for details.

name	value	initial value	min	max	vary	expression
m_PU_SG_PU_1_n1_1	1.78461575	1.784615753610107	1.0000e-04	5.00000000	False	m_PU_SG_PU_1_1_n1
m_PU_SG_PU_1_1_n1	1.78461575	1.784615753610107	1.0000e-04	5.00000000	True
m_PU_SG_PU_3_n1_1	0.80290495	0.802904951323892	1.0000e-04	5.00000000	False	m_PU_SG_PU_3_1_n1
m_PU_SG_PU_3_1_n1	0.80290495	0.802904951323892	1.0000e-04	5.00000000	True
m_SU_SG_SU_1_0_0	2.68606212	2.686062120382649	1.0000e-04	5.00000000	True
m_SU_SG_SU_3_0_0	1.10915311	1.109153108617096	1.0000e-04	5.00000000	True
p_PU_SG_PU_1_n1_1	-0.86104140	-0.8610414024232179	-3.14159265	3.14159265	False	p_PU_SG_PU_1_1_n1
p_PU_SG_PU_1_1_n1	-0.86104140	-0.8610414024232179	-3.14159265	3.14159265	True
p_PU_SG_PU_3_n1_1	-3.12044446	-3.1204444620772467	-3.14159265	3.14159265	False	p_PU_SG_PU_3_1_n1
p_PU_SG_PU_3_1_n1	-3.12044446	-3.1204444620772467	-3.14159265	3.14159265	True
p_SU_SG_SU_1_0_0	2.61122920	2.611229196458127	-3.14159265	3.14159265	True
p_SU_SG_SU_3_0_0	-0.07867828	-0.07867827542158025	-3.14159265	3.14159265	True



*** Setup demo fitting workspace OK.

Run fits

For this notebook, we’ll either (a) run a batch of fits or (b) load sample data.

With the current codebase, running multiple fits will default to using the same basis set, and output results sequentially to the main self.data dictionary. (Note this currently runs in serial.)

(a) Run a batch

[2]:

import time

start = time.time()

[3]:

for n in range(0,1000):
    data.randomizeParams()
    data.fit()

[4]:

end = time.time()
print(end - start)

19807.788444519043

[10]:

# We now have 100 fit results
data.data.keys()

[10]:

dict_keys(['orb6', 'orb5', 'ADM', 'pol', 'subset', 'sim', 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99])

[5]:

# Quick data dump
# TODO: better save routine (json/h5).

import pickle
with open('dataDump_1000fitTests_150621.pickle', 'wb') as handle:
    pickle.dump(data.data, handle, protocol=pickle.HIGHEST_PROTOCOL)

(b) Load a batch of fit runs

Load sample data for analysis instead of running fits. Note this can be run minimally without the full setup routines above, using the commented-out cell below to init a blank object.

(The demo file(s) are available in demos/fitting.)

[1]:

# If running from scratch, create a blank object first
# # Init blank object
from pemtk.fit.fitClass import pemtkFit
data = pemtkFit()

*** ePSproc installation not found, setting for local copy.

[2]:

# Load sample dataset
# Full path to the file may be required here, in demos/fitting
import pickle

with open('dataDump_100fitTests_10t_randPhase_130621.pickle', 'rb') as handle:
    data.data = pickle.load(handle)

[3]:

data.data.keys()

[3]:

dict_keys(['orb6', 'orb5', 'ADM', 'pol', 'subset', 'sim', 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99])

Exploring a fit result

Each result contains a set of fit results:

[10]:

nFit = 11
data.data[nFit].keys()

[10]:

dict_keys(['AFBLM', 'residual', 'results'])

Here ‘results’ is an lmFit object, containing various outputs, including the final paramter set and fit statistics, which can be inspected directly. (See the basic demo notebook for more.)

[11]:

data.data[nFit]['results']

[11]:

Fit Statistics

fitting method	leastsq
# function evals	493
# data points	195
# variables	8
chi-square	1.3656e-04
reduced chi-square	7.3029e-07
Akaike info crit.	-2747.48342
Bayesian info crit.	-2721.29942

Variables

name	value	standard error	relative error	initial value	min	max	vary	expression
m_PU_SG_PU_1_n1_1	1.58289364	0.00420180	(0.27%)	0.9969358394309723	1.0000e-04	5.00000000	False	m_PU_SG_PU_1_1_n1
m_PU_SG_PU_1_1_n1	1.58289364	0.00420180	(0.27%)	0.9969358394309723	1.0000e-04	5.00000000	True
m_PU_SG_PU_3_n1_1	1.15075298	0.00591749	(0.51%)	0.6501789011457593	1.0000e-04	5.00000000	False	m_PU_SG_PU_3_1_n1
m_PU_SG_PU_3_1_n1	1.15075298	0.00591749	(0.51%)	0.6501789011457593	1.0000e-04	5.00000000	True
m_SU_SG_SU_1_0_0	2.71014401	0.00245323	(0.09%)	0.5256290247418078	1.0000e-04	5.00000000	True
m_SU_SG_SU_3_0_0	1.04870406	0.00589304	(0.56%)	0.3431948628999326	1.0000e-04	5.00000000	True
p_PU_SG_PU_1_n1_1	1.22200072	28873.8615	(2362835.06%)	0.22504503901184436	-3.14159265	3.14159265	False	p_PU_SG_PU_1_1_n1
p_PU_SG_PU_1_1_n1	1.22200072	28873.8615	(2362835.07%)	0.22504503901184436	-3.14159265	3.14159265	True
p_PU_SG_PU_3_n1_1	-1.09376359	28873.8574	(2639862.74%)	0.2384618778656229	-3.14159265	3.14159265	False	p_PU_SG_PU_3_1_n1
p_PU_SG_PU_3_1_n1	-1.09376359	28873.8574	(2639862.73%)	0.2384618778656229	-3.14159265	3.14159265	True
p_SU_SG_SU_1_0_0	-2.28911439	28873.8581	(1261354.97%)	0.2118753318675165	-3.14159265	3.14159265	True
p_SU_SG_SU_3_0_0	0.90233526	28873.8543	(3199903.15%)	0.5407367643463352	-3.14159265	3.14159265	True

Correlations (unreported correlations are < 0.100)

p_PU_SG_PU_3_1_n1	p_SU_SG_SU_1_0_0	1.0000
p_PU_SG_PU_1_1_n1	p_PU_SG_PU_3_1_n1	1.0000
p_PU_SG_PU_1_1_n1	p_SU_SG_SU_1_0_0	1.0000
p_PU_SG_PU_1_1_n1	p_SU_SG_SU_3_0_0	1.0000
p_PU_SG_PU_3_1_n1	p_SU_SG_SU_3_0_0	1.0000
p_SU_SG_SU_1_0_0	p_SU_SG_SU_3_0_0	1.0000
m_PU_SG_PU_1_1_n1	m_PU_SG_PU_3_1_n1	-0.9519
m_SU_SG_SU_1_0_0	m_SU_SG_SU_3_0_0	-0.8360
m_PU_SG_PU_1_1_n1	p_SU_SG_SU_3_0_0	-0.5513
m_PU_SG_PU_1_1_n1	p_PU_SG_PU_1_1_n1	-0.5513
m_PU_SG_PU_1_1_n1	p_PU_SG_PU_3_1_n1	-0.5513
m_PU_SG_PU_1_1_n1	p_SU_SG_SU_1_0_0	-0.5513
m_PU_SG_PU_3_1_n1	p_SU_SG_SU_3_0_0	0.5195
m_PU_SG_PU_3_1_n1	p_PU_SG_PU_1_1_n1	0.5195
m_PU_SG_PU_3_1_n1	p_PU_SG_PU_3_1_n1	0.5195
m_PU_SG_PU_3_1_n1	p_SU_SG_SU_1_0_0	0.5195
m_SU_SG_SU_3_0_0	p_SU_SG_SU_1_0_0	-0.3544
m_SU_SG_SU_3_0_0	p_PU_SG_PU_1_1_n1	-0.3544
m_SU_SG_SU_3_0_0	p_PU_SG_PU_3_1_n1	-0.3544
m_SU_SG_SU_3_0_0	p_SU_SG_SU_3_0_0	-0.3544
m_SU_SG_SU_1_0_0	p_PU_SG_PU_1_1_n1	0.3413
m_SU_SG_SU_1_0_0	p_SU_SG_SU_1_0_0	0.3413
m_SU_SG_SU_1_0_0	p_PU_SG_PU_3_1_n1	0.3413
m_SU_SG_SU_1_0_0	p_SU_SG_SU_3_0_0	0.3413
m_PU_SG_PU_3_1_n1	m_SU_SG_SU_1_0_0	-0.1329

The best fit results are set in an Xarray, keyed by AFBLM.

[12]:

data.data[nFit]['AFBLM']

For further analysis & batch results, see the “analysis” notebook.

Versions

[13]:

import scooby
scooby.Report(additional=['epsproc', 'pemtk', 'xarray', 'jupyter'])

[13]:

Tue Jun 15 12:25:36 2021 Eastern Daylight Time
OS	Windows	CPU(s)	32	Machine	AMD64
Architecture	64bit	RAM	63.9 GB	Environment	Jupyter
Python 3.7.3 (default, Apr 24 2019, 15:29:51) [MSC v.1915 64 bit (AMD64)]
epsproc	1.3.0-dev	pemtk	0.0.1	xarray	0.15.0
jupyter	Version unknown	numpy	1.19.2	scipy	1.3.0
IPython	7.12.0	matplotlib	3.3.1	scooby	0.5.6
Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191125 for Intel(R) 64 architecture applications

[23]:

# Check current Git commit for local ePSproc version
from pathlib import Path
!git -C {Path(ep.__file__).parent} branch
!git -C {Path(ep.__file__).parent} log --format="%H" -n 1

fatal: cannot change to '{Path(ep.__file__).parent}': No such file or directory
fatal: cannot change to '{Path(ep.__file__).parent}': No such file or directory

[15]:

# Check current remote commits
!git ls-remote --heads git://github.com/phockett/ePSproc
# !git ls-remote --heads git://github.com/phockett/epsman

16cfad26e658b740f267baa89d1550336b0134bf        refs/heads/dev
82d12cf35b19882d4e9a2cde3d4009fe679cfaee        refs/heads/master
69cd89ce5bc0ad6d465a4bd8df6fba15d3fd1aee        refs/heads/numba-tests
ea30878c842f09d525fbf39fa269fa2302a13b57        refs/heads/revert-9-master

[21]:

# Check current Git commit for local PEMtk version
import pemtk
from pathlib import Path
!git -C {Path(pemtk.__file__).parent} branch
!git -C {Path(pemtk.__file__).parent} log --format="%H" -n 1

* master
be904921ea27af085682088bb4460770c6ca55e6

[22]:

# Check current remote commits
!git ls-remote --heads git://github.com/phockett/PEMtk
# !git ls-remote --heads git://github.com/phockett/epsman

cdca35025d0790f5c32d714a942bbed7796f7aa6        refs/heads/master