def get_anchor_flderr_param(i, train, N, model, mode, fd):
# return both the projection and rom fld relative errro
import numpy as np
import pandas as pd
import reader
import re
dirs = '../ra_' + str(train) + '_theta_90/' + model[
i] + '_parameter_ra_' + str(train) + '/mrelerr/'
dirs = '../re' + str(train) + '/' + model[i] + '_parameter_Re' + str(
train) + '/mrelerr_h1/'
if model[i] == 'l-rom':
filename = 'param_list_' + mode + '.dat'
angle = np.loadtxt(dirs + filename)
filename = 'rom_relerr_N' + str(
N[i]) + '_' + fd[i] + '_' + mode + '.dat'
fldrelerr_rom = np.loadtxt(dirs + filename)
filename = 'proj_relerr_N' + str(N[i]) + '_' + mode + '.dat'
fldrelerr_proj = np.loadtxt(dirs + filename)
elif model[i] == 'l-rom-df':
fname = 'mrelerr_N' + str(N[i]) + '_0' + fd[i] + '.csv'
else:
fname = 'mrelerr_N' + str(N[i]) + '.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
if all(x in f for x in ['Ra' + str(train), model[i], str(N[i])]):
# if 'Ra'+str(train) in re.split('[/_]', f):
# if 'Re'+str(train) in re.split('[/_]', f):
filename = f
data = pd.read_csv(filename)
return [data[i].tolist() for i in data.columns]
def get_FOM_mtfluc(P_test):
import sys
sys.path.append('/Users/bigticket0501/Developer/PyMOR/code/plot_helpers/')
import re
import numpy as np
from reader import find_files
fom = []
for i, test in enumerate(P_test):
fname = 'tmtfluc'
filenames = find_files(fname, '../')
filenames = find_files(fname, '../../')
for f in filenames:
if 'Ra_'+str(test) in re.split('[/]', f):
filename = f
data = np.loadtxt(filename)
fom.append(data)
return fom
def get_opterri_param(P_test,
P_train,
N,
P_test_anchor,
model,
mode,
fd,
scaled=''):
import numpy as np
import pandas as pd
import reader
# get erri with theta_g at each anchor points
erri_his = []
for i, train in enumerate(P_train):
if model[i] == 'l-rom':
fname = 'dual_norm_N' + str(
N[i]) + '_' + fd[i].strip('0') + scaled + '_' + mode + '.dat'
elif model[i] == 'l-rom-df':
fname = 'dual_norm_N' + str(N[i]) + '_0' + fd[i] + scaled + '.csv'
else:
fname = 'dual_norm_N' + str(N[i]) + scaled + '.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
if all(x in f for x in ['Ra' + str(train), model[i], str(N[i])]):
# if 'Ra'+str(train) in re.split('[/_]', f):
# if 'Re'+str(train) in re.split('[/_]', f):
filename = f
data = pd.read_csv(filename)
param, erri = [data[i].tolist() for i in data.columns]
idx = [param.index(i) for i in P_test]
param = np.asarray(param)[idx]
erri = np.asarray(erri)[idx]
erri_his.append(erri)
erri_comb = np.array(erri_his)
erri_opt = []
P_train = list(P_train)
for i, test in enumerate(P_test):
index = P_train.index(P_test_anchor[i])
erri_opt.append(erri_comb[index, i])
return param, erri_comb, erri_opt
def get_anchor_mtke(i, train, N, model, mode, fd, feature):
import pandas as pd
import reader
if model[i] == 'l-rom':
fname = feature+'_N'+str(N[i])+'_'+fd[i].strip('0')+'_'+mode+'.csv'
elif model[i] == 'l-rom-df':
fname = feature+'_N'+str(N[i])+'_0'+fd[i]+'.csv'
else:
fname = feature+'_N'+str(N[i])+'.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
if all(x in f for x in ['Ra'+str(train), model[i], str(N[i])]):
filename = f
data = pd.read_csv(filename)
param = data['Ra'].tolist()
data1 = data[feature].tolist()
return param, data1
def get_anchor_qoi_param(i, train, N, model, mode, fd):
import numpy as np
import pandas as pd
import re
import reader
dirs = '../ra_' + str(train) + '_theta_90/' + model[
i] + '_parameter_ra_' + str(train) + '/nu/'
if model[i] == 'l-rom':
fname = 'mnurelerr_N' + str(
N[i]) + '_' + fd[i].strip('0') + '_' + mode + '.csv'
elif model[i] == 'l-rom-df':
fname = 'mnurelerr_N' + str(N[i]) + '_0' + fd[i] + '.csv'
else:
fname = 'mnurelerr_N' + str(N[i]) + '.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
# if 'Ra'+str(train) in re.split('[/_]', f):
if all(x in f for x in ['Ra' + str(train), model[i], str(N[i])]):
filename = f
data = pd.read_csv(filename)
param = data['Ra'].tolist()
merr = data['mnurelerr'].tolist()
if model[i] == 'l-rom':
fname = 'mnu_N' + str(
N[i]) + '_' + fd[i].strip('0') + '_' + mode + '.csv'
elif model[i] == 'l-rom-df':
fname = 'mnu_N' + str(N[i]) + '_0' + fd[i] + '.csv'
else:
fname = 'mnu_N' + str(N[i]) + '.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
# if 'Ra'+str(train) in re.split('[/_]', f):
if all(x in f for x in ['Ra' + str(train), model[i], str(N[i])]):
filename = f
data = pd.read_csv(filename)
m = data['mnu'].tolist()
if model[i] == 'l-rom':
fname = 'std_nu_N' + str(
N[i]) + '_' + fd[i].strip('0') + '_' + mode + '.csv'
elif model[i] == 'l-rom-df':
fname = 'std_nu_N' + str(N[i]) + '_0' + fd[i] + '.csv'
else:
fname = 'std_nu_N' + str(N[i]) + '.csv'
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
for f in filenames:
# if 'Ra'+str(train) in re.split('[/_]', f):
if all(x in f for x in ['Ra' + str(train), model[i], str(N[i])]):
filename = f
data = pd.read_csv(filename)
std = data['std_nu'].tolist()
return param, merr, m, std
def get_FOM_nu_1st2nd(P_test):
import sys
sys.path.append('/Users/bigticket0501/Developer/PyMOR/code/plot_helpers/')
import re
import pandas as pd
from reader import find_files
fom_m_list = []
fom_sd_list = []
for i, test in enumerate(P_test):
fname = 'nus_mom.csv'
filenames = find_files(fname, '../')
filenames = find_files(fname, '../../')
for f in filenames:
if 'Ra_'+str(test) in re.split('[/]', f):
filename = f
data = pd.read_csv(filename)
fom_mean = data['mean']
fom_std = data[' std']
fom_m_list.append(fom_mean)
fom_sd_list.append(fom_std)
return fom_m_list, fom_sd_list
# filename = '../../../fom_nuss/nuss_fom_'+str(test)
# filename = '../fom_nuss/nus_fom_'+str(test)
# data = mypostpro.read_nuss(filename)
# if (str(test) == '10000'):
# data[:, 2] = data[:, 2]/40
# else:
# data[:, 2] = data[:, 2]
# idx1 = mypostpro.find_nearest(data[:, 0], 0)
# idx2 = mypostpro.find_nearest(data[:, 0], 1000)
# nuss_fom = data[idx1:idx2, :]
# avgidx1 = mypostpro.find_nearest(data[:, 0], 501)
# fom_mean = mypostpro.cmean(nuss_fom[avgidx1:idx2], 2)
# fom_var = mypostpro.cvar(nuss_fom[avgidx1:idx2], fom_mean, 2)
# fom_sd = mypostpro.csd(nuss_fom[avgidx1:idx2], fom_mean, 2)
fname = 'nus_mom.csv'
filenames = find_files(fname, '../')
for f in filenames:
if 'Ra_' + str(test) in re.split('[/]', f):
filename = f
data = pd.read_csv(filename)
fom_mean = data['mean']
fom_std = data[' std']
fom_m_list.append(fom_mean)
fom_sd_list.append(fom_std)
fig1, ax1 = plt.subplots(1, tight_layout=True)
fig2, ax2 = plt.subplots(1, tight_layout=True)
fig3, ax3 = plt.subplots(1, tight_layout=True)
fig4, ax4 = plt.subplots(1, tight_layout=True)
def conv_compare(argv):
import yaml
import os
import re
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from figsetup.style import style
from figsetup.text import text
from aux.create_dir import create_dir
sys.path.append('/home/pht2/Developer/PyROM/code/plot_helpers/')
import reader
style(1)
text()
models = []
for i in range(len(argv) - 1):
models.append(argv[i])
feature = argv[-1]
fig, ax = plt.subplots(1, tight_layout=True)
for model in models:
dir1 = model + '_reproduction'
with open('reproduction.yaml') as f:
info = yaml.load(f, Loader=yaml.FullLoader)
for key, value in info['parameters'].items():
al = '_'.join([str(key), str(value)])
dir1 = '_'.join([dir1, al])
if model == 'l-rom':
fd = info['perc'].replace('p', '.')
fd = str(int(float(fd) * 100))
solver = model.upper() + ' with ' + str(
fd) + ' percentage filtered'
fname = feature + '_' + info['perc'] + '.csv'
elif model == 'l-rom-df':
fd = info['fwidth'].replace('p', '.')
solver = model.upper() + r' with filter width $\delta=$ ' + str(fd)
fname = feature + '_' + info['fwidth'] + '.csv'
else:
solver = model.upper()
fname = feature + '.csv'
tpath = reader.find_files(fname, './')
print(tpath)
for f in tpath:
if model in re.split('[/_]', f):
fn = f
# tpath = os.path.join(dir1, feature, fname)
data = pd.read_csv(fn)
ax.plot(data.iloc[:, 0], data.iloc[:, 1], '-o', label=solver)
ax.legend(loc=0)
anc_lb = []
for key, value in info['parameters'].items():
if key == 'theta':
anc_lb.append('\\' + str(key) + '^*_g=' + str(value))
else:
anc_lb.append(str(key) + '^*=' + str(value))
anc_lb = ', '.join(anc_lb)
if feature == 'mrelerr_h1':
title = 'Relative error in the predicted mean flow at ' + '$' + anc_lb + '$'
ax.set(
xlabel=r'$N$',
ylabel=
r'$\frac{\|\langle \bf{u} - \bf{\tilde{u}} \rangle\|_{H^1}}{\|\langle \bf{u} \rangle\|_{H^1}}$',
ylim=[1e-3, 1],
title=title)
ax.set_yscale('log')
elif feature == 'mtke':
title = 'Predicted mean TKE at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$', ylabel=r'$\langle TKE \rangle_g$', title=title)
ax.set_yscale('log')
mtke_fom = np.loadtxt('../qoi/tmtke')
fom_params = {'c': 'k', 'marker': 'o', 'label': 'FOM'}
ax.plot(data.iloc[:, 0], mtke_fom * np.ones(len(data.iloc[:, 0])),
**fom_params)
elif feature == 'dual_norm':
title = 'Dual norm at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$',
ylabel=r'$\triangle$',
title=title,
ylim=[1e-4, 1])
ax.set_yscale('log')
elif feature == 'mtfluc':
title = 'Predicted mean fluctuation in temperature at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$',
ylabel=r'$\langle T_{fluc} \rangle_s$',
title=title)
mtfluc_fom = np.loadtxt('../qoi/tmtfluc')
fom_params = {'c': 'k', 'marker': 'o', 'label': 'FOM'}
ax.plot(data.iloc[:, 0], mtfluc_fom * np.ones(len(data.iloc[:, 0])),
**fom_params)
elif feature == 'mnu':
filename = './fom/nus_mom.csv'
fom = pd.read_csv(filename).to_numpy()
title = 'Predicted mean Nu at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$', ylabel=r'$\langle Nu \rangle_s$', title=title)
fom_params = {'c': 'k', 'marker': 'o', 'label': 'FOM'}
ax.plot(data.iloc[:, 0], fom[0][0] * np.ones(len(data.iloc[:, 0])),
**fom_params)
elif feature == 'stdnu':
filename = './fom/nus_mom.csv'
fom = pd.read_csv(filename).to_numpy()
title = 'Predicted std(Nu) at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$', ylabel=r'Std(Nu)', title=title)
fom_params = {'c': 'k', 'marker': 'o', 'label': 'FOM'}
ax.plot(data.iloc[:, 0], fom[0][1] * np.ones(len(data.iloc[:, 0])),
**fom_params)
elif feature == 'mnu_err':
title = 'Relative error in mean Nu at ' + '$' + anc_lb + '$'
ax.set(xlabel=r'$N$', title=title)
ax.set_yscale('log')
ax.legend(loc=0)
tdir = './compare/'
create_dir(tdir)
fig.savefig(tdir + feature + '_conv_compare.png')
return
def get_ncand_param(P_test, P_train, ncand, N, model, mode, fd, scaled=''):
import numpy as np
import pandas as pd
import reader
import re
P_test_anchor = []
# Find out the nearest anchor(s) point to each testing parameter
for i, test in enumerate(P_test):
erris = []
near_anch = []
# distance between two parameters is used
tmp = abs(test - P_train)
print(test, tmp, P_train, len(set(tmp)))
if (len(set(tmp)) < len(P_train)):
ncand = 3
elif (len(set(tmp)) == len(P_train)):
ncand = 2
for j in range(ncand):
# find out the index corresponding to the closest anchor point
if 0 in tmp:
minidx = np.argmin(tmp)
near_anch.append(P_train[minidx])
else:
minidx = np.argmin(tmp)
near_anch.append(P_train[minidx])
tmp[minidx] = 1e8
# do not revisit the same anchor again
# aflag = test in P_train
# print(aflag, test, P_train)
# if aflag:
# pass
# else:
# tmp[minidx] = 1e8
if model[minidx] == 'l-rom':
fname = 'erri_N' + str(N[minidx]) + '_' + fd[minidx].strip(
'0') + scaled + '_' + mode + '.dat'
elif model[minidx] == 'l-rom-df':
fname = 'dual_norm_N' + str(
N[minidx]) + '_0' + fd[minidx] + scaled + '.csv'
else:
fname = 'erri_N' + str(
N[minidx]) + scaled + '_' + mode + '.dat'
fname = 'dual_norm_N' + str(N[minidx]) + scaled + '.csv'
# dual_norm = np.loadtxt(dirs+fname)
# data = np.array(dual_norm).astype(np.float64)
filenames = reader.find_files(fname, '../')
filenames = reader.find_files(fname, '../../')
# print(['Ra'+str(int(P_train[minidx])), model[minidx], str(N[minidx])])
for f in filenames:
if all(x in f for x in [
'Ra' + str(int(P_train[minidx])), model[minidx],
str(N[minidx])
]):
# if 'Ra'+str(int(P_train[minidx])) in re.split('[/_]', f) and model[minidx] in re.split('[/_]', f):
# if 'Re'+str(int(P_train[minidx])) in re.split('[/_]', f):
filename = f
data = pd.read_csv(filename)
param, erri = [data[k].tolist() for k in data.columns]
idx = param.index(test)
erris.append(erri[idx])
idx = erris.index(min(erris))
print(i, test, near_anch, idx)
P_test_anchor.append(near_anch[idx])
return P_test_anchor