Skip to content
Snippets Groups Projects
Commit 19fd9789 authored by Xavier Adriaens's avatar Xavier Adriaens
Browse files

up plot scripts

parent 36bb1710
Branches
No related tags found
No related merge requests found
Show whitespace changes
Inline Side-by-side
input/thesis_part1/plots/directional_plot_partial.py
+ 4
− 4
View file @ 19fd9789
......@@ -10,7 +10,7 @@ import matplotlib.pyplot as plt
from matplotlib import rc
plt.rcParams['ytick.right'] = plt.rcParams['ytick.labelright'] = True
plt.rcParams['ytick.left'] = plt.rcParams['ytick.labelleft'] = False
rc('text.latex', preamble=r'\usepackage{amsmath} \usepackage{amssymb}')
rc('text.latex', preamble=r'\usepackage{amsmath} \usepackage{amssymb} \renewcommand{\familydefault}{\sfdefault} \usepackage{sansmathfonts}')
rc('font',**{'family':'serif','sans-serif':['Helvetica'],'size':10})
rc('text', usetex=True)
H = 3.0;
......@@ -31,7 +31,7 @@ def plot(data):
plt.plot(x[:,data.idx_param],x[:,2*data.n_param],color='k',marker='o',markersize=3,linewidth=0)
plt.xlabel(r'$'+data.var+'$')
plt.ylabel(r'$\mathcal{J}('+data.var+')$')
plt.ylabel(r'$J('+data.var+')$')
plt.grid()
plt.savefig(prepath+data.path+data.filename+'_misfit.eps')
plt.close()
......@@ -65,7 +65,7 @@ def plot(data):
plt.grid()
#plt.legend(loc=3,ncol=1);
plt.xlabel(r'$'+data.var+'$')
plt.ylabel(r'$D_{'+data.var+'} \mathcal{J}('+data.var+')$')
plt.ylabel(r'$D_{'+data.var+'} J('+data.var+')$')
plt.savefig(prepath+data.path+data.filename+'_dj.eps')
plt.close()
......@@ -96,7 +96,7 @@ def plot(data):
plt.grid()
#plt.legend(loc=3,ncol=1);
plt.xlabel(r'$'+data.var+'$')
plt.ylabel(r'$D^2_{'+data.var+'} \mathcal{J}('+data.var+')$')
plt.ylabel(r'$D^2_{'+data.var+'} J('+data.var+')$')
plt.savefig(prepath+data.path+data.filename+'_d2j.eps')
plt.close()
......
inversion_plot_comp_part2.py 0 → 100644
+ 479
− 0
View file @ 19fd9789
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Wed Aug 3 16:57:28 2022
@author: xavier
"""
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import rc
plt.rcParams['ytick.right'] = plt.rcParams['ytick.labelright'] = False
plt.rcParams['ytick.left'] = plt.rcParams['ytick.labelleft'] = True
rc('text.latex', preamble=r'\usepackage{amsmath} \usepackage{amssymb} \renewcommand{\familydefault}{\sfdefault} \usepackage{sansmathfonts}')
rc('font',**{'family':'serif','sans-serif':['Helvetica'],'size':10})
rc('text', usetex=True)
lw = 1.5
H = 2.0;
L = 3.0;
"""
#CS1: IP
j0 = 35.6633
n_comp = 4
path = ['extended/US_lBFGS/','extended/S_lBFGS/','extended/SST_lBFGS/','extended/SSM_lBFGS/']
filename = ['paper2_extended_inversion_lbfgs','paper2_extended_inversion_lbfgs_w','paper2_extended_inversion_lbfgs_w_stab','paper2_extended_inversion_lbfgs_w_smooth']
lab = ['unweight','weight','stab','smooth']
glob = ['linesearch','linesearch','linesearch','linesearch']
order = [1, 1, 1, 1]
col = ['k','k', 'k', 'k']
mrk = ['']*4
lin = [':','-.','--','-']
max_forward = 80
Δmax_forward = 10
n_freq = 1
"""
"""
#CS1: GD
j0 = 35.6633
n_comp = 7
path = ['extended/SST_GD_LS/','extended/SST_GD_TRA/','extended/SST_GD_BTRB/','extended/SST_GD_BTRC/','extended/SST_GD_BTRAR/','extended/SST_GD_BTRBR/','extended/SST_GD_BTRCR/']
filename = ['paper2_extended_sst_gd_ls','paper2_extended_sst_gd_trA','paper2_extended_sst_gd_trB','paper2_extended_sst_gd_trC','paper2_extended_sst_gd_trAR','paper2_extended_sst_gd_trBR','paper2_extended_sst_gd_trCR']
lab = ['LS','TR-P(A)','TR-P(B)','TR-P(C)','TR-R(A)','TR-R(B)','TR-R(C)']
glob = ['linesearch'] + ['trust-region'] * 3 + ['trust-region-retro'] * 3
order = [1] * 7
col = ['k'] + ['C0','C1','C2'] + ['C3','C4','C5']
mrk = ['']*7
#lin = ['-']*7
lin = [(0, (4, 4)), (0, (4, 4)), (0, (4, 4)),(4, (4, 4)), (0, (4, 8)), (4, (4, 8)), (8, (4, 8))]
max_forward = 360
Δmax_forward = 40
n_freq = 1
"""
"""
#CS1: l-BFGS
j0 = 35.6633
n_comp = 7
path = ['extended/SST_lBFGS/','extended/SST_lBFGS_TRA/','extended/SST_lBFGS_TRB/','extended/SST_lBFGS_TRC/','extended/SST_lBFGS_TRAR/','extended/SST_lBFGS_TRBR/','extended/SST_lBFGS_TRCR/']
filename = ['paper2_extended_sst_lbfgs_ls','paper2_extended_sst_lbfgs_trA','paper2_extended_sst_lbfgs_trB','paper2_extended_sst_lbfgs_trC','paper2_extended_sst_lbfgs_trAR','paper2_extended_sst_lbfgs_trBR','paper2_extended_sst_lbfgs_trCR']
lab = ['LS','TR-P(A)','TR-P(B)','TR-P(C)','TR-R(A)','TR-R(B)','TR-R(C)']
glob = ['linesearch'] + ['trust-region'] * 3 + ['trust-region-retro'] * 3
order = [1.5] * 7
col = ['k'] + ['C0','C1','C2'] + ['C3','C4','C5']
mrk = ['']*7
#lin = ['-'] * 7
lin = [(0, (4, 4)), (0, (4, 4)), (0, (4, 8)),(0, (4, 4)), (4, (4, 4)), (4, (4, 8)), (8, (4, 8))]
max_forward = 80
Δmax_forward = 10
n_freq = 1
"""
"""
#CS1: Newton
j0 = 35.6633
n_comp = 14
path = ['extended/SST_NCG/','extended/SST_NSCGA/','extended/SST_NSCGB/','extended/SST_NSCGC/','extended/SST_NSCGAR/','extended/SST_NSCGBR/','extended/SST_NSCGCR/','extended/SST_NCGGN/','extended/SST_NSCGAGN/','extended/SST_NSCGBGN/','extended/SST_NSCGCGN/','extended/SST_NSCGARGN/','extended/SST_NSCGBRGN/','extended/SST_NSCGCRGN/']
filename = ['paper2_extended_sst_ncg','paper2_extended_sst_nscgA','paper2_extended_sst_nscgB','paper2_extended_sst_nscgC','paper2_extended_sst_nscgAr','paper2_extended_sst_nscgBr','paper2_extended_sst_nscgC','paper2_extended_sst_ncggn','paper2_extended_sst_nscgAgn','paper2_extended_sst_nscgBgn','paper2_extended_sst_nscgCgn','paper2_extended_sst_nscgArgn','paper2_extended_sst_nscgBrgn','paper2_extended_sst_nscgCrgn']
lab = ['ncg','nscgA','nscgB','nscgC','nscgAr','nscgBr','nscgCr','ncggn','nscgAgn','nscgBgn','nscgCgn','nscgArgn','nscgBrgn','nscgCrgn']
glob = (['linesearch'] + ['trust-region'] * 3 + ['trust-region-retro'] * 3) * 2
order = [2] * 14
col = ['k', 'C0','C1','C2', 'C3', 'C4', 'C5'] * 2
mrk = ['']*14
lin = [(0,(4,4))]*14
lin[0] = (0,(4,4))
#lin[1] = (0,(4,4))
#lin[2] = (2,(4,4))
#lin[3] = (4,(4,4))
#lin[0] = (4,(4,4))
lin[1] = (2,(4,4))
lin[2] = (0,(4,4))
lin[3] = (4,(4,4))
lin[4] = (0,(4,4))
lin[5] = (4,(4,4))
lin[4+7] = (0,(4,4))
lin[5+7] = (4,(4,4))
lin[0+7] = (4,(4,4))
lin[2+7] = (0,(4,4))
lin[3+7] = (4,(4,4))
max_forward = 180
Δmax_forward = 20
n_freq = 1
"""
"""
#CS2: IP
j0 = 209.852
n_comp = 4
path = ['extended/concrete3/US_LBFGS_LS/','extended/concrete3/S_LBFGS_LS/','extended/concrete3/SST_LBFGS_LS/','extended/concrete3/SSM_LBFGS_LS/']
filename = ['paper2_concrete3_us_lbfgs_ls','paper2_concrete3_s_lbfgs_ls','paper2_concrete3_sst_lbfgs_ls','paper2_concrete3_ssm_lbfgs_ls']
lab = ['unweight','weight','stab','smooth']
glob = ['linesearch','linesearch','linesearch','linesearch']
order = [1, 1, 1, 1]
col = ['k','k', 'k', 'k']
mrk = ['']*4
lin = [':','-.','--','-']
lin[2] = (4,(4,4))
max_forward = 200+10
Δmax_forward = 25
n_freq = 1
"""
"""
#CS2: GD
j0 = 209.852
n_comp = 3
path = ['extended/concrete3/SSM_GD_LS/','extended/concrete3/SSM_GD_BTRB/','extended/concrete3/SSM_GD_BTRBR/']
filename = ['paper2_concrete3_ssm_gd_ls','paper2_concrete3_ssm_gd_trB','paper2_concrete3_ssm_gd_trBR']
lab = ['LS','TR-P(B)','TR-R(B)']
glob = ['linesearch'] + ['trust-region'] * 1 + ['trust-region-retro'] * 1
order = [1] * 3
col = ['k'] + ['C1'] + ['C4']
mrk = ['']*3
lin = ['-']*3
max_forward = 800
Δmax_forward = 100
n_freq = 1
"""
"""
#CS2: l-BFGS
j0 = 209.852
n_comp = 3
path = ['extended/concrete3/SSM_LBFGS_LS/','extended/concrete3/SSM_LBFGS_TRB/','extended/concrete3/SSM_LBFGS_TRBR/']
filename = ['paper2_concrete3_ssm_lbfgs_ls','paper2_concrete3_ssm_lbfgs_trB','paper2_concrete3_ssm_lbfgs_trBR']
lab = ['LS','TR-P(B)','TR-R(B)']
glob = ['linesearch'] + ['trust-region'] * 1 + ['trust-region-retro'] * 1
order = [1.5] * 3
col = ['k'] + ['C1'] + ['C4']
mrk = ['']*3
lin = ['-']*3
max_forward = 200
Δmax_forward = 25
n_freq = 1
"""
"""
#CS2: Newton
j0 = 209.852
n_comp = 6
path = ['extended/concrete3/SSM_NCG/','extended/concrete3/SSM_NSCGB/','extended/concrete3/SSM_NSCGBR/','extended/concrete3/SSM_NCGGN/','extended/concrete3/SSM_NSCGBGN/','extended/concrete3/SSM_NSCGBRGN/']
filename = ['paper2_concrete3_ssm_ncg','paper2_concrete3_ssm_nscgB','paper2_concrete3_ssm_nscgBR','paper2_concrete3_ssm_ncggn','paper2_concrete3_ssm_nscgBgn','paper2_concrete3_ssm_nscgBRgn']
lab = ['FN-LS','FN-TR-P(B)','FN-TR-R(B)','GN-LS','GN-TR-P(B)','GN-TR-R(B)']
glob = ['linesearch','trust-region','trust-region-retro'] * 2
order = [2.] * 6
col = ['k','C1','C4']*2
mrk = ['o']*6
lin = ['-']*6
offset = [1]*3 + [0]*3
max_forward = 600
Δmax_forward = 75
n_freq = 1
"""
"""
#CS3: IP
j0 = 0.0328706
n_comp = 4
path = ['TC3/LBFGS_LS_SST5/','TC3/LBFGS_LS_SST2/','TC3/LBFGS_LS_SST6/','TC3/LBFGS_LS_SST3/']
filename = ['cross_inversion_sst5_lbfgs_ls','cross_inversion_sst2_lbfgs_ls','cross_inversion_sst6_lbfgs_ls','cross_inversion_sst3_lbfgs_ls']
lab = [r'$\beta = 2$',r'$\beta = 3$',r'$\beta = 4$',r'$\beta = 5$']
glob = ['linesearch'] * 4
order = [1] * 4
col = ['k', '#696969', '#A9A9A9', '#D3D3D3']
mrk = [''] * 4
lin = ['-'] * 4
max_forward = 900
Δmax_forward = 80
n_freq = 1
"""
#CS3: L-BFGS
j0 = 0.0328706
n_comp = 3
path = ['TC3/LBFGS_LS_SST2/','TC3/LBFGS_TRB_SST2/','TC3/LBFGS_TRBR_SST2/']
filename = ['cross_inversion_sst2_lbfgs_ls','cross_inversion_sst2_lbfgs_trB','cross_inversion_sst2_lbfgs_trBR']
lab = ['LS','TR-P(B)','TR-R(B)']
glob = ['linesearch'] + ['trust-region'] * 1 + ['trust-region-retro'] * 1
order = [1.5] * 3
col = ['k'] + ['C1'] + ['C4']
mrk = ['']*3
lin = ['-']*3
max_forward = 900
Δmax_forward = 80
offset = [1]*6
n_freq = 1
"""
#CS3: Newton
j0 = 0.0328706
n_comp = 6
path = ['TC3/NCG_SST2/','TC3/NSCGB_SST2/','TC3/NSCGBR_SST2/','TC3/NCGGN_SST2/','TC3/NSCGBGN_SST2/','TC3/NSCGBRGN_SST2/']
filename = ['cross_inversion_sst2_ncg','cross_inversion_sst2_nscgB','cross_inversion_sst2_nscgBR','cross_inversion_sst2_ncggn','cross_inversion_sst2_nscgBgn','cross_inversion_sst2_nscgBRgn']
lab = ['FN-LS','FN-TR-P(B)','FN-TR-R(B)','GN-LS','GN-TR-P(B)','GN-TR-R(B)']
glob = ['linesearch','trust-region','trust-region-retro'] * 2
order = [2.] * 6
col = ['k','C1','C4']*2
mrk = ['o']*6
lin = ['-']*6
max_forward = 900
Δmax_forward = 80
offset = [1]*6
n_freq=1
"""
"""
LOAD
"""
jcfs = []
p_ratiocfs = []
r_ratiocfs = []
n_itcf = []
n_scalecf = []
n_itcfs = []
for i in range(0,n_comp):
n_scalef = [[]]*n_freq;
jfs = [[]]*n_freq;
p_ratiofs = [[]]*n_freq;
r_ratiofs = [[]]*n_freq;
n_itfs = [[]]*n_freq;
n_itf = [0]*n_freq
scalefs = [[]]*n_freq;
for f in range(0,n_freq):
n_scalef[f] = np.loadtxt(path[i]+filename[i]+'_globalminimum_historyg'+str(f)+'.csv',delimiter=';',max_rows=1,skiprows=3,dtype=np.integer)[1];
scalefs[f] = np.loadtxt(path[i]+filename[i]+'_globalminimum_historyg'+str(f)+'.csv',delimiter=';',usecols=2,max_rows=np.int(n_scalef[f]),skiprows=6,dtype=np.float);
jfs[f] = [[]]*n_scalef[f];
p_ratiofs[f] = [[]]*n_scalef[f];
r_ratiofs[f] = [[]]*n_scalef[f];
n_itfs[f] = np.zeros(n_scalef[f],np.integer);
line = 3
for s in range(0,n_scalef[f]):
n_itfs[f][s] = np.loadtxt(path[i]+filename[i]+'_localminimum_historyg'+str(f)+'.csv',delimiter=';',usecols = 1, max_rows=1,skiprows=line,dtype=np.integer);
n_itf[f] = n_itf[f]+n_itfs[f][s]
line=line+2
jfs[f][s] = np.loadtxt(path[i]+filename[i]+'_localminimum_historyg'+str(f)+'.csv',delimiter=';',max_rows=np.int(n_itfs[f][s]+1),skiprows=line,dtype=np.float)[:,0] / j0;
if glob[i] == 'trust-region' or glob[i] == 'trust-region-retro':
p_ratiofs[f][s] = np.loadtxt(path[i]+filename[i]+'_localminimum_historyg'+str(f)+'.csv',delimiter=';',max_rows=np.int(n_itfs[f][s]+1),skiprows=line,dtype=np.float)[:,5+offset[i]];
r_ratiofs[f][s] = np.loadtxt(path[i]+filename[i]+'_localminimum_historyg'+str(f)+'.csv',delimiter=';',max_rows=np.int(n_itfs[f][s]+1),skiprows=line,dtype=np.float)[:,6+offset[i]];
r_ratiofs[f][s][p_ratiofs[f][s] < 0] = np.nan;
line=line+n_itfs[f][s]+1+1
jcfs.append(jfs)
p_ratiocfs.append(p_ratiofs)
r_ratiocfs.append(r_ratiofs)
n_itcf.append(n_itf)
n_itcfs.append(n_itfs)
n_scalecf.append(n_scalef)
it_linecf = [[]*n_freq]*n_comp;
it_descentcf = [[]*n_freq]*n_comp;
bnd_descentcf = [[]*n_freq]*n_comp;
n_forwardcf = [[]*n_freq]*n_comp;
successcf = [[]*n_freq]*n_comp;
negativecf = [[]*n_freq]*n_comp;
etacf = [[]*n_freq]*n_comp;
for i in range(0,n_comp):
it_linef = [[]]*n_freq
it_descentf = [[]]*n_freq
bnd_descentf = [[]]*n_freq
n_forwardf = [[]]*n_freq
successf = [[]]*n_freq
negativef = [[]]*n_freq
etaf = [[]]*n_freq
for f in range(0,n_freq):
n_forwardf[f] = np.zeros(n_itcf[i][f],np.integer)
if glob[i] == 'linesearch':
it_linef[f] = np.loadtxt(path[i]+filename[i]+'_linesearch_historyg'+str(f)+'.csv',delimiter=';',usecols = 5, skiprows=1,dtype=np.integer)
n_forwardf[f] += it_linef[f]
if order[i] == 1.5 and (glob[i] == 'trust-region' or glob[i] == 'trust-region-retro') :
successf[f] = np.zeros(n_itcf[i][f],np.integer)
bnd_descentf[f] = np.zeros(n_itcf[i][f],np.integer)
line = 2
for n in range(0,n_itcf[i][f]):
bnd_descentf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=4,max_rows=1,skiprows=line,dtype=np.integer);
successf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=0,max_rows=1,skiprows=line,dtype=np.integer);
line=line+2
if order[i] == 2:
it_descentf[f] = np.zeros(n_itcf[i][f],np.integer)
successf[f] = np.zeros(n_itcf[i][f],np.integer)
etaf[f] = np.zeros(n_itcf[i][f],np.float)
negativef[f] = np.zeros(n_itcf[i][f],np.integer)
bnd_descentf[f] = np.zeros(n_itcf[i][f],np.integer)
line = 3
for n in range(0,n_itcf[i][f]):
it_descentf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=3,max_rows=1,skiprows=line,dtype=np.integer);
if glob[i] == 'trust-region' or glob[i] == 'trust-region-retro':
bnd_descentf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=6,max_rows=1,skiprows=line,dtype=np.integer);
etaf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=1,max_rows=1,skiprows=line,dtype=np.float);
successf[f][n] = np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=0,max_rows=1,skiprows=line,dtype=np.integer);
line=line+it_descentf[f][n]+1
negativef[f][n] = (np.loadtxt(path[i]+filename[i]+'_descent_historyg'+str(f)+'.csv',delimiter=';',usecols=1,max_rows=1,skiprows=line,dtype=np.float)<0);
line=line+2
n_forwardf[f] += 2*it_descentf[f]
if glob[i] == 'trust-region-retro':
n_forwardf[f] += 2 #retrospective Hessian-vector product
n_forwardf[f] += 2
n_forwardf[f] = np.cumsum(n_forwardf[f])
n_forwardf[f] = np.insert(n_forwardf[f],0,0)
n_forwardcf[i] = n_forwardf
it_descentcf[i] = it_descentf
bnd_descentcf[i] = bnd_descentf
it_linecf[i] = it_linef
successcf[i] = successf
etacf[i] = etaf
negativecf[i] = negativef
"""
DATA CONVERGENCE PLOT
"""
fig = [[]]*n_freq
for f in range(0,n_freq):
fig[f]=plt.figure(figsize=(2*L,H),tight_layout=False);
plt.subplots_adjust(top=0.95,right=0.92,bottom=0.13,left=0.08)
plt.ticklabel_format(style='sci', axis='y', scilimits=(0,0))
plt.grid(zorder=0)
plt.xlim([0,max_forward])
plt.xticks(np.arange(0,max_forward+1,Δmax_forward))
#plt.ylim([5e-4,2])
plt.ylim([5e-4,2])
plt.yscale('log')
for i in (0,1,2):#range(0,n_comp):
n_it=0
lines = [[]]*n_scalecf[i][f]
plt.plot(n_forwardcf[i][f][n_it],jcfs[i][f][0][0],color=col[i],marker='o',zorder=3,markersize=3)
for s in range(0,n_scalecf[i][f]):
lines[s] = plt.plot(n_forwardcf[i][f][range(n_it,n_it+n_itcfs[i][f][s]+1)],jcfs[i][f][s][0:n_itcfs[i][f][s]+1],color=col[i],linestyle=lin[i],zorder=2,label=lab[i],marker=mrk[i],markersize=5.,linewidth=lw)
plt.plot(n_forwardcf[i][f][n_it+n_itcfs[i][f][s]],jcfs[i][f][s][n_itcfs[i][f][s]],color=col[i],marker='x',zorder=3,markersize=3.)
n_it=n_it+n_itcfs[i][f][s]
n_it = 0
#plt.legend()
#plt.savefig('convergence_comp'+str(f)+'.eps')
"""
HISTORY CONVERGENCE PLOT (Retrospective/prospective ratio)
"""
fig = [[]]*n_freq
for f in range(0,n_freq):
fig[f]=plt.figure(figsize=(2*L,H / 2.),tight_layout=False);
plt.subplots_adjust(top=0.95,right=0.92,bottom=0.23,left=0.08)
plt.ticklabel_format(style='sci', axis='y', scilimits=(0,0))
plt.grid(zorder=0)
plt.yticks(np.linspace(0,1.5,4))
plt.ylim([0.0,1.75])
plt.xticks(np.arange(0,23,2))
plt.xlim([1,23.5])
#plt.xticks(np.arange(0,46,5))
#plt.xlim([1,45.5])
for i in (1,2):
if glob[i] == 'trust-region':
plt.plot(range(1,n_itcfs[i][f][0]+1),p_ratiocfs[i][f][0][1:],color=col[i],linestyle=lin[i],zorder=2,label=r'$\rho_{\text{p}}$',linewidth=lw)
if glob[i] == 'trust-region-retro':
plt.plot(range(1,n_itcfs[i][f][0]+1),r_ratiocfs[i][f][0][1:],color=col[i],linestyle=lin[i],zorder=2,label=r'$\rho_{\text{r}}$',linewidth=lw)
#plt.legend()
#plt.savefig('ratio_nscgrgn.eps')
"""
HISTORY CONVERGENCE PLOT (forcing sequence)
"""
fig = [[]]*n_freq
for f in range(0,n_freq):
fig[f]=plt.figure(figsize=(2*L,H / 2.),tight_layout=False);
plt.subplots_adjust(top=0.93,right=0.92,bottom=0.23,left=0.08)
plt.grid(zorder=0)
plt.ylim([0.3,1.])
plt.yticks(np.arange(0.4,1.01,0.2))
#plt.xticks(np.arange(0,23,2))
#plt.xlim([0,22.5])
plt.xticks(np.arange(0,46,5))
plt.xlim([0,45.5])
for i in (3,):#range(0,n_comp,1):
plt.plot(range(0,n_itcfs[i][f][0]),np.sqrt(etacf[i][f][:]),color=col[i],linestyle=lin[i],zorder=2,label=r'$\rho_{\text{p}}$',linewidth=lw)
#plt.legend()
#plt.savefig('eta_comp.eps')
"""
HISTORY CONVERGENCE PLOT (inner iteration per outer iteration)
"""
fig = [[]]*n_freq
for f in range(0,n_freq):
fig[f]=plt.figure(figsize=(2*L,H / 2.),tight_layout=False);
plt.subplots_adjust(top=0.93,right=0.92,bottom=0.23,left=0.08)
plt.ticklabel_format(style='plain', axis='y', scilimits=(0,0))
plt.grid(zorder=0)
#plt.xticks(np.arange(0,23,2))
#plt.yticks(np.arange(0,15,4))
#plt.xlim([0,22.5])
#plt.ylim([0,14.5])
plt.xticks(np.arange(0,46,5))
plt.yticks(np.arange(0,41,10))
plt.xlim([0,45.5])
plt.ylim([0,40.])
#plt.xticks(np.arange(0,23,2))
#plt.yticks(np.arange(0,101,25))
#plt.xlim([0,22.5])
#plt.ylim([0,100])
for i in (0,1,2): #range(0,n_comp):
plt.plot(range(0,n_itcfs[i][f][0]),it_descentcf[i][f][:],color=col[i],linestyle=lin[i],zorder=2,linewidth=lw)
#plt.savefig('it_comp.eps')
"""
MODEL CONVERGENCE PLOT
"""
#To investigate !!!!
errorcfs = []
for i in range(0,n_comp):
errorc = np.loadtxt(path[i]+filename[i]+'_error.csv',delimiter=';')[:,0];
errorcfs.append(errorc)
#plt.plot(n_forwardcf[i][0],errorc,color=col[i],linestyle=lin[i],zorder=2,label=lab[i])
plt.plot(jcfs[i][0][0],errorc,color=col[i],linestyle=lin[i],zorder=2,label=lab[i])
plt.legend()
plt.xscale('log')
plt.gca().invert_xaxis()
"""
STATS
"""
print(''.rjust(10), 'Wave (tot.)'.rjust(20), 'Outer it. (tot.)'.rjust(20), 'Inner it. (mean)'.rjust(20), 'Failure (tot.)'.rjust(20), 'Bnd reached (tot.)'.rjust(20), 'Success (tot.)'.rjust(20), 'Negative curv. (tot.)'.rjust(20))
for i in range(0,n_comp):
outter = np.sum(n_itcf[i])
inner = 0
failure = 0
boundary = 0
success = 0
negative = 0
call = 0
for f in range(0,n_freq):
call += n_forwardcf[i][f][-1]
inner += np.sum(it_descentcf[i][f])
if glob[i] == 'linesearch':
failure += np.count_nonzero(it_linecf[i][f])
elif glob[i] == 'trust-region' or glob[i] == 'trust-region-retro':
for s in range(0,n_scalecf[i][f]):
failure += n_itcfs[i][f][s] - np.count_nonzero(np.diff(jcfs[i][f][s]))
boundary += np.sum(bnd_descentcf[i][f])
if order[i] == 1.5 and (glob[i] == 'trust-region' or glob[i] == 'trust-region-retro') :
success += np.sum(successcf[i][f])
if order[i] == 2:
success += np.sum(successcf[i][f])
negative += np.sum(negativecf[i][f])
print(lab[i].rjust(10), str(call).rjust(20), str(outter).rjust(20), str(inner/outter).rjust(20), str(failure/outter).rjust(20), str(boundary/outter).rjust(20), str(success/outter).rjust(20), str(negative/outter).rjust(20) )
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment