def main(functionals, func, category='Aminoindan', bins=600, reg_m=10**2):
#####################
#### SETUP START ####
#####################
############################################ Directory with KF files
kf_dir = r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals"
#####################
#### SETUP END ####
#####################
func_best = {
comp_funcs.wasserstein_distance: min,
comp_funcs.wasserstein_distance_unbalanced: min,
comp_funcs.l2: min,
comp_funcs.diagonality: max,
comp_funcs.bhattacharyya: min,
comp_funcs.correlation: min,
comp_funcs.chi_square: min,
comp_funcs.kl_divergence: min
}
#get kffiles
kff = [f for f in os.listdir(kf_dir)]
kff = list(
filter(lambda x: functionals[0] in x or functionals[1] in x, kff))
kff_of_cat = [kf_dir + '\\' + f for f in kff if f.startswith(category)]
#get the spectra
ir1 = [
ir.get_spectrum_from_kf(f, width=50, n=bins) for f in kff_of_cat
if functionals[0] in f
]
ir2 = [
ir.get_spectrum_from_kf(f, width=50, n=bins) for f in kff_of_cat
if functionals[1] in f
]
d = func(ir1, ir2, reg_m=reg_m)
return d.astype(float)
kff_list = [kf_dir + '\\' + f for f in kff if f.startswith(category)] # kff_list = [ # r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E8_DFTB3_DFTB.rkf", # r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E7_LDA_DFT.t21", # ] kff_list = [ r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E12_DFTB3_DFTB.rkf", r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E22_LDA_DFT.t21", ] #get the spectra ir1 = [ir.get_spectrum_from_kf(f,width=50,n=bins) for f in kff_list if functionals[0] in f] ir2 = [ir.get_spectrum_from_kf(f,width=50,n=bins) for f in kff_list if functionals[1] in f] peaksa = [ir.get_freqs_intens(f) for f in kff_list if functionals[0] in f] freqsa, intensa = [list(c) for c in zip(*peaksa)] peaksa = [list(zip(freqsa[i], intensa[i])) for i in range(len(freqsa))] peaksb = [ir.get_freqs_intens(f) for f in kff_list if functionals[1] in f] freqsb, intensb = [list(c) for c in zip(*peaksb)] peaksb = [list(zip(freqsb[i], intensb[i])) for i in range(len(freqsb))] def setup(): global animation_folder animation_folder = os.getcwd() + r'\animation\\'
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI5_DFT.t21",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI6_DFT.t21"]
# kf_dftb = [r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI1_DFTB.rkf",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI2_DFTB.rkf",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI3_DFTB.rkf",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI4_DFTB.rkf",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI5_DFTB.rkf",
# r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\conf_unopt\AI6_DFTB.rkf"]
#get the spectra
ir_dft = [ir.get_spectrum_from_kf(f,width=50) for f in kf_dft]
ir_dftb = [ir.get_spectrum_from_kf(f,width=50) for f in kf_dftb]
#normalize spectra
ir_dft = [i/np.sum(i) for i in ir_dft]
ir_dftb = [i/np.sum(i) for i in ir_dftb]
#wasserstein distances
Y, X = np.meshgrid(np.linspace(0,1,600), np.linspace(0,1,600))
C = abs(Y-X)**2
print('Case (Balanced Wasserstein)')
d = np.zeros((len(ir_dft), len(ir_dftb)))
for i, a in enumerate(ir_dft/np.sum(ir_dft)):
bottom_line = Border(bottom=Side(style='thin'))
right_line = Border(right=Side(style='thin'))
both_line = Border(right=Side(style='thin'), bottom=Side(style='thin'))
all_line = Border(top=Side(style='thick'), left=Side(style='thick'), right=Side(style='thick'), bottom=Side(style='thick'))
bold_huge = Font(bold=False, size=30)
bold_large = Font(bold=False, size=16)
bold_small = Font(bold=False, size=11)
#write each category to a new sheet
for cat in categories:
#generate the spectra
ir_dft = [ir.get_spectrum_from_kf(f,width=50) for f in kff_by_cat[cat] if f.endswith('.t21')]
ir_dftb = [ir.get_spectrum_from_kf(f,width=50) for f in kff_by_cat[cat] if f.endswith('.rkf')]
#setup new sheet
ws = wb.create_sheet(cat)
wb.active = ws
ws['A1'] = cat
ws['A1'].font = bold_huge
data_col_offset = 5 #distance from the left side of the sheet where to write data
n = len(ir_dft)
#iterate over comparison functions
for i, func in enumerate(funcs):
right_line = Border(right=Side(style='thin'))
both_line = Border(right=Side(style='thin'), bottom=Side(style='thin'))
all_line = Border(top=Side(style='thick'),
left=Side(style='thick'),
right=Side(style='thick'),
bottom=Side(style='thick'))
bold_huge = Font(bold=False, size=30)
bold_large = Font(bold=False, size=16)
bold_small = Font(bold=False, size=11)
for i_cat, s_name, cat in zip(range(len(categories)), sheet_names, categories):
#get the spectra
ir_dft = [
ir.get_spectrum_from_kf(f, width=50) for f in kff_by_cat[cat]
if f.endswith('.t21')
]
ir_dftb = [
ir.get_spectrum_from_kf(f, width=50) for f in kff_by_cat[cat]
if f.endswith('.rkf')
]
# print(np.sum(ir_dft), np.sum(ir_dftb))
#setup sheet
ws = wb.create_sheet(s_name)
wb.active = ws
ws['A1'] = s_name
ws['A1'].font = bold_huge
import modules.plot as plot
import modules.barycenter as bc
import modules.ir as ir
import os
# kf_path = r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\[21-05](13-05-2020)\\"
# kfs = [kf_path + d for d in os.listdir(kf_path)]
kfs = [
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_1_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_1_DFTB.rkf",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_2_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_2_DFTB.rkf",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_3_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_3_DFTB.rkf",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_4_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_4_DFTB.rkf",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_5_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_5_DFTB.rkf",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_6_DFT.t21",
r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\second_set\Aminoindan_CONF_6_DFTB.rkf",
]
for kf in kfs:
spec = ir.get_spectrum_from_kf(kf, xlim=(0, 4000), n=1000)
plot.plot_hist(spec,
title=kf.split('\\')[-1][:-4],
axislabels=('Wavenumber ($cm^{-1}$)', 'Intensity'),
xlim=(0, 4000),
invert_xaxis=True)
#choose two histograms a and b:
# a = hist.gaussian(400, 0.7, 0.1, 0)
# a = hist.gaussian(400, 0.7, 0.1, 0) + 0.1 * hist.gaussian(400, 0.5,0.03)
# a = hist.gaussian(400, 0.2, 0.06) + hist.gaussian(400, 0.5, 0.06)*2 + hist.gaussian(400, 0.8, 0.06)
# a = hist.slater(400, 0.5, 30, 0)
# a = hist.from_func(400, lambda x: 1-x**2)
# a = hist.from_func(400, func)
# a = hist.from_func(400, lambda x: 1-x,0)
# a = hist.from_func(400, lambda x: np.cos(x*6*3.14)+1)
# a = hist.from_func(400, lambda x: x)
# a = hist.dirac_delta(400, 0.5)
# a = hist.from_func(400, lambda x: ((x-0.5)*10)**4) + 3*hist.gaussian(400, 0.7, 0.1)
# r = jobs.DFTJob('l-alanine', job_name='l-alanine_DFT').run(); a = ir.get_spectrum(r, xlim=(0,2000), n=400)
a = ir.get_spectrum_from_kf(
r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\Aminoindan_CONF_4_LDA_DFT.t21",
xlim=(0, 4000),
n=400)
# b = hist.gaussian(400, 0.3, 0.1, 0)
# b = hist.gaussian(400, 0.8, 0.05, 0) + hist.gaussian(400, 0.2, 0.05, 0)
# b = hist.gaussian(400, 0.5, 0.05, 0)
# b = hist.gaussian(400, 0.2, 0.06)*2 + hist.gaussian(400, 0.5, 0.06) + hist.gaussian(400, 0.8, 0.06)*2
# b = hist.from_func(400, lambda x: np.sin(x*10*3.14)+1)
# b = hist.from_func(400, func2)
# b = hist.from_func(400, lambda x: x,0)
# b = hist.from_func(400, lambda x: x**0)
# b = hist.slater(400, 0.8, 30, 0)
# b = hist.from_func(400, lambda x: np.cos(x*5*3.14)+1)
# r = jobs.DFTBJob('l-alanine', job_name='l-alanine_DFT').run(); b = ir.get_spectrum(r, xlim=(0,2000), n=400)
b = ir.get_spectrum_from_kf(
r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\Aminoindan_CONF_4_DFTB3_freq_DFTB.rkf",
def main(functionals, save_name=None, sub_dir='', bins=600, reg_m=10**2):
#####################
#### SETUP START ####
#####################
############################################ Comparison functions
funcs = [
# comp_funcs.wasserstein_distance,
# comp_funcs.wasserstein_distance_unbalanced,
comp_funcs.freq_int_wasserstein,
# comp_funcs.l2,
# comp_funcs.diagonality,
# comp_funcs.bhattacharyya,
# comp_funcs.correlation,
# comp_funcs.chi_square,
# comp_funcs.kl_divergence,
]
############################################ Colour map for data colouring
colour_map = cmap.WhiteGreen()
############################################ Directory with KF files
kf_dir = r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals"
# kf_dir = r"C:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\random_set"
############################################ Categories to split the KF files into
# categories = ['Aminoindan_CONF_', 'ISO34_E', 'ISO34_P', 'SCONF']
# categories = ['Aminoindan_CONF_', 'ISO34_E', 'SCONF']
# categories = ['Aminoindan_CONF_']
categories = ['']
############################################ Sheet names
sheet_names = categories
############################################ Functionals to split the KF files into
# functionals = ['OLYP_DFT', 'LDA_DFT', 'DFTB3_DFTB', 'DFTB3_freq_DFTB']
# functionals = ['DFTB3_freq_DFTB', 'LDA_DFT']
############################################ Path to save xl file to
if save_name is None:
save_name = kf_dir.split('\\')[-1] + f'_comparisons_{functionals[0]}_{functionals[1]}'
else:
save_name = kf_dir.split('\\')[-1] + '_' + save_name
# save_name = kf_dir.split('\\')[-1] + '_comparisons'
#############################################
#####################
#### SETUP END ####
#####################
assert len(categories) == len(sheet_names)
func_best = {comp_funcs.wasserstein_distance: min,
comp_funcs.wasserstein_distance_unbalanced: min,
comp_funcs.freq_int_wasserstein: min,
comp_funcs.l2: min,
comp_funcs.diagonality: max,
comp_funcs.bhattacharyya: min,
comp_funcs.correlation: min,
comp_funcs.chi_square: min,
comp_funcs.kl_divergence: min}
#get kffiles
kff = [f for f in os.listdir(kf_dir)]
kff = list(filter(lambda x: functionals[0] in x or functionals[1] in x, kff))
#separate into categories
kff_by_cat = {}
for cat in categories:
kff_by_cat[cat] = [kf_dir + '\\' + f for f in kff if f.startswith(cat)]
names_by_cat = {}
for cat in categories:
names_by_cat[cat] = {functionals[0]:[], functionals[1]:[]}
for f in kff:
name = f
if name.startswith(cat):
name = name[len(cat):][:-4]
if functionals[0] in name:
names_by_cat[cat][functionals[0]].append(name)
elif functionals[1] in name:
names_by_cat[cat][functionals[1]].append(name)
#start writing
out_file = f"excel_files/{sub_dir}/{save_name}_{bins}.xlsx"
try:
os.mkdir(f"excel_files/{sub_dir}")
except:
pass
wb = xl.Workbook()
ws0 = wb.active
bottom_line = Border(bottom=Side(style='thin'))
right_line = Border(right=Side(style='thin'))
both_line = Border(right=Side(style='thin'), bottom=Side(style='thin'))
all_line = Border(top=Side(style='thick'), left=Side(style='thick'), right=Side(style='thick'), bottom=Side(style='thick'))
bold_huge = Font(bold=False, size=30)
bold_large = Font(bold=False, size=16)
bold_small = Font(bold=False, size=11)
for i_cat, s_name, cat in zip(range(len(categories)), sheet_names, categories):
#get the spectra
ir1 = [ir.get_spectrum_from_kf(f,width=50,n=bins) for f in kff_by_cat[cat] if functionals[0] in f]
ir2 = [ir.get_spectrum_from_kf(f,width=50,n=bins) for f in kff_by_cat[cat] if functionals[1] in f]
peaksa = [ir.get_freqs_intens(f) for f in kff_by_cat[cat] if functionals[0] in f]
freqsa, intensa = [list(c) for c in zip(*peaksa)]
peaksa = [list(zip(freqsa[i], intensa[i])) for i in range(len(freqsa))]
peaksb = [ir.get_freqs_intens(f) for f in kff_by_cat[cat] if functionals[1] in f]
freqsb, intensb = [list(c) for c in zip(*peaksb)]
peaksb = [list(zip(freqsb[i], intensb[i])) for i in range(len(freqsb))]
#setup sheet
ws = wb.create_sheet(s_name)
wb.active = ws
ws['A1'] = s_name
ws['A1'].font = bold_huge
data_col_offset = 5
n = len(ir1)
for i, func in enumerate(funcs):
#setup new matching table
fn = str(func).split()[1] #func name
print(cat + '_' + fn)
ws.cell(row=i*(n+3)+3, column=1).value = fn.upper()
ws.cell(row=i*(n+3)+3, column=1).font = bold_large
ws.cell(row=i*(n+3)+4, column=1).value = 'Best'
ws.cell(row=i*(n+3)+4, column=1).border = bottom_line
ws.cell(row=i*(n+3)+4, column=1).font = bold_small
ws.cell(row=i*(n+3)+4, column=2).value = 'Error'
ws.cell(row=i*(n+3)+4, column=2).border = bottom_line
ws.cell(row=i*(n+3)+4, column=2).font = bold_small
ws.cell(row=i*(n+3)+4, column=3).value = 'Error rel'
ws.cell(row=i*(n+3)+4, column=3).border = bottom_line
ws.cell(row=i*(n+3)+4, column=3).font = bold_small
ws.cell(row=i*(n+3)+4, column=data_col_offset).value = rf'{functionals[0]}\{functionals[1]}'
ws.cell(row=i*(n+3)+4, column=data_col_offset).border = both_line
ws.cell(row=i*(n+3)+4, column=data_col_offset).font = bold_small
# kwargs = {'reg_m': np.linspace(-1,2,7)[i_cat]}
d = func(ir1, ir2, reg_m=reg_m, peaksa=peaksa, peaksb=peaksb)
if func_best[func] is min:
dnorm = (d-d.min())/(d.max()-d.min())
if func_best[func] is max:
dnorm = 1-(d-d.min())/(d.max()-d.min())
c = colour_map.get_hex_colours(dnorm)
for j, row in enumerate(d):
try:
ws.cell(row=i*(n+3)+5+j, column=data_col_offset).value = names_by_cat[cat][functionals[0]][j]
ws.cell(row=i*(n+3)+5+j, column=data_col_offset).border = right_line
ws.cell(row=i*(n+3)+5+j, column=1).value = func_best[func](row)
ws.cell(row=i*(n+3)+5+j, column=2).value = row[j] - func_best[func](row)
ws.cell(row=i*(n+3)+5+j, column=3).value = f'{(row[j] - func_best[func](row))/row[j] * 100:.2f}%'
for k, el in enumerate(row):
ws.cell(row=i*(n+3)+4, column=data_col_offset+1+k).value = names_by_cat[cat][functionals[1]][k]
ws.cell(row=i*(n+3)+4, column=data_col_offset+1+k).border = bottom_line
ws.cell(row=i*(n+3)+5+j, column=data_col_offset+1+k).value = el
ws.cell(row=i*(n+3)+5+j, column=data_col_offset+1+k).fill = PatternFill(start_color=c[j,k].decode("utf-8"), end_color=c[j,k].decode("utf-8"), fill_type = "solid")
if el == func_best[func](row):
ws.cell(row=i*(n+3)+5+j, column=data_col_offset+1+k).border = all_line
except:
raise
wb.save(out_file)
del wb['Sheet']
wb.save(out_file)
def main(functionals,
func,
category='Aminoindan',
bins=600,
reg_m=10**2,
**kwargs):
#####################
#### SETUP START ####
#####################
############################################ Directory with KF files
kf_dir = r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals"
#####################
#### SETUP END ####
#####################
func_best = {
comp_funcs.wasserstein_distance: min,
comp_funcs.wasserstein_distance_unbalanced: min,
comp_funcs.freq_int_wasserstein: min,
comp_funcs.l2: min,
comp_funcs.diagonality: max,
comp_funcs.bhattacharyya: min,
comp_funcs.correlation: min,
comp_funcs.chi_square: min,
comp_funcs.kl_divergence: min
}
#get kffiles
kff = [f for f in os.listdir(kf_dir)]
kff = list(
filter(lambda x: functionals[0] in x or functionals[1] in x, kff))
kff_of_cat = [kf_dir + '\\' + f for f in kff if f.startswith(category)]
#get the spectra
ir1 = [
ir.get_spectrum_from_kf(f, width=50, n=bins) for f in kff_of_cat
if functionals[0] in f
]
ir2 = [
ir.get_spectrum_from_kf(f, width=50, n=bins) for f in kff_of_cat
if functionals[1] in f
]
peaksa = [
ir.get_freqs_intens(f) for f in kff_of_cat if functionals[0] in f
]
freqsa, intensa = [list(c) for c in zip(*peaksa)]
peaksa = [list(zip(freqsa[i], intensa[i])) for i in range(len(freqsa))]
peaksb = [
ir.get_freqs_intens(f) for f in kff_of_cat if functionals[1] in f
]
freqsb, intensb = [list(c) for c in zip(*peaksb)]
peaksb = [list(zip(freqsb[i], intensb[i])) for i in range(len(freqsb))]
print(peaksa)
d = func(ir1, ir2, reg_m=reg_m, peaksa=peaksa, peaksb=peaksb, **kwargs)
return d.astype(float)
plt.legend()
plt.show()
# plot.plot_hists(specs, ('mol1 - DFTB3', 'mol2 - LDA'), line=False, scatter=True)
sys.exit()
specs = tuple([s / np.sum(s) for s in specs])
res = sink.sinkhorn(*specs, converge_thresh=converge_thresh, max_iter=1000)
plot.plot_sink_results(res)
plot.plot_hists(specs, ('mol1 - DFTB3', 'mol2 - LDA'))
specs = [
ir.get_spectrum_from_kf(
r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E22_DFTB3_DFTB.rkf",
xlim=(0, 4000),
n=600),
ir.get_spectrum_from_kf(
r"D:\Users\Yuman\Desktop\Programmeren\bachelorproject\scripting\RUNS\#KFFiles\functionals\ISO34_E22_LDA_DFT.t21",
xlim=(0, 4000),
n=600),
]
specs = tuple([s / np.sum(s) for s in specs])
res = sink.sinkhorn(*specs, converge_thresh=converge_thresh, max_iter=1000)
plot.plot_sink_results(res)
plot.plot_hists(specs, ('mol1 - DFTB3', 'mol1 - LDA'))
sys.exit()
Y, X = np.meshgrid(np.linspace(0, 1, len(a)), np.linspace(0, 1, len(b)))
C = cost_fn(X, Y)