In [1]:
import pandas as pd
from pandas.io.json import json_normalize #package for flattening json in pandas df
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
import json
import os
%matplotlib inline
In [2]:
os.chdir("./runs/approx/main2/")
origin_path = os.getcwd()
runs_id = os.listdir("./")
runs_id = [int(item) for item in runs_id]
runs_id = sorted(runs_id)
In [3]:
df_list = []
for run_id in runs_id:
os.chdir("./"+str(run_id))
f_in = open('out.json', 'r')
run_info = json.load(f_in)
f_in.close()
a = json_normalize(run_info)
df_list.append(a)
os.chdir(origin_path)
res1 = pd.concat(df_list,ignore_index=True)
In [7]:
res_correct = res1
res_correct_2d = res_correct[res_correct['Model info.dim'] == 2]
res_correct_3d = res_correct[res_correct['Model info.dim'] == 3]
In [28]:
res_correct_2d[res_correct_2d['Out info.MSE'] < 0.428e-2]
Out[28]:
In [30]:
res_correct_3d[res_correct_3d['Out info.MSE'] < 0.4e-1]
Out[30]:
In [9]:
matplotlib.rcParams['text.usetex'] = True
matplotlib.rcParams['text.latex.unicode'] = True
matplotlib.rcParams['text.latex.preamble'] = [r'\usepackage{amsmath}']
In [11]:
res_correct = res_correct_2d
m_sig_all = np.array(res_correct['Model info.n_sig'].tolist())
mse_all = np.array(res_correct['Out info.MSE'].tolist())
mse_means = []
std_errs = []
h = 4
for i in range(50):
ds = res_correct[0+h*i:h+h*i]
mse_l = np.array(ds['Out info.MSE'].tolist())
mse_means.append(np.mean(mse_l))
std_errs.append(np.std(mse_l))
my_set = set(m_sig_all)
m_sig = sorted( list(my_set) )
figure = plt.figure(figsize=(12,8))
axes = figure.add_subplot (1, 1, 1)
plt.grid(True)
plt.title(r'$\sigma_{\text{sol}}^2$ vs number of neurons. 2D task', fontsize=25)
plt.xlabel('Number of neurons', fontsize=25)
plt.ylabel(r'$\sigma_{\text{sol}}^2$', fontsize=25)
plt.scatter(m_sig_all, mse_all, label = r'$\sigma_{\text{sol}}^2$ for all points', marker = "D",s=40)
plt.plot(m_sig, mse_means, color='black', marker='x', linestyle='dashed', linewidth=3, markersize=16, label = r'Mean $\sigma_{\text{sol}}^2$')
plt.errorbar(m_sig, mse_means, yerr=std_errs, ecolor='r', lw=2, capsize=15, mew = 3, zorder=3, label = r'Std.error of $\sigma_{\text{sol}}^2$', linestyle='None')
axes.set_yscale ('log')
plt.legend(loc=1, prop={'size': 20})
Out[11]:
In [14]:
res_correct = res_correct_3d
m_sig_all = np.array(res_correct['Model info.n_sig'].tolist())
mse_all = np.array(res_correct['Out info.MSE'].tolist())
mse_means = mse_all
m_sig = m_sig_all
#mse_means = []
#std_errs = []
#h = 4
#for i in range(50):
# ds = res_correct[0+h*i:h+h*i]
# mse_l = np.array(ds['Out info.MSE'].tolist())
# mse_means.append(np.mean(mse_l))
# std_errs.append(np.std(mse_l))
#my_set = set(m_sig_all)
#m_sig = sorted( list(my_set) )
figure = plt.figure(figsize=(12,8))
axes = figure.add_subplot (1, 1, 1)
plt.grid(True)
plt.title(r'$\sigma_{\text{sol}}^2$ vs number of neurons. 3D task', fontsize=25)
plt.xlabel('Number of neurons', fontsize=25)
plt.ylabel(r'$\sigma_{\text{sol}}^2$', fontsize=25)
#plt.scatter(m_sig_all, mse_all, label = r'$\sigma_{\text{sol}}^2$ for all points', marker = "D",s=40)
plt.plot(m_sig, mse_means, color='black', marker='x', linestyle='dashed', linewidth=3, markersize=16, label = r'$\sigma_{\text{sol}}^2$')
#plt.errorbar(m_sig, mse_means, yerr=std_errs, ecolor='r', lw=2, capsize=15, mew = 3, zorder=3, label = r'Std.error of $\sigma_{\text{sol}}^2$', linestyle='None')
axes.set_yscale ('log')
plt.legend(loc=1, prop={'size': 20})
Out[14]:
In [ ]:
In [ ]:
In [18]:
m_sig_all = np.array(res_correct['Model info.n_sig'].tolist())
mse_all = np.array(res_correct['Out info.MSE'].tolist())
mse_means = []
std_errs = []
h = 4
for i in range(50):
ds = res_correct[0+h*i:h+h*i]
mse_l = np.array(ds['Out info.MSE'].tolist())
mse_means.append(np.mean(mse_l))
std_errs.append(np.std(mse_l))
my_set = set(m_sig_all)
m_sig = sorted( list(my_set) )
figure = plt.figure(figsize=(12,8))
axes = figure.add_subplot (1, 1, 1)
plt.grid(True)
plt.title(r'$\sigma_{\text{sol}}^2$ vs number of neurons. One-dimensional task', fontsize=25)
plt.xlabel('Number of neurons', fontsize=25)
plt.ylabel(r'$\sigma_{\text{sol}}^2$', fontsize=25)
plt.scatter(m_sig_all, mse_all, label = r'$\sigma_{\text{sol}}^2$ for all points', marker = "D",s=40)
plt.plot(m_sig, mse_means, color='black', marker='x', linestyle='dashed', linewidth=3, markersize=16, label = r'Mean $\sigma_{\text{sol}}^2$')
plt.errorbar(m_sig, mse_means, yerr=std_errs, ecolor='r', lw=2, capsize=15, mew = 3, zorder=3, label = r'Std.error of $\sigma_{\text{sol}}^2$', linestyle='None')
axes.set_yscale ('log')
plt.legend(loc=1, prop={'size': 20})
Out[18]:
In [ ]:
In [ ]:
In [ ]:
m_sig_all = np.array(res_correct['Model info.n_sig'].tolist())
J_all = np.array(res_correct['Out info.J_fin'].tolist())
J_means = []
std_errs = []
h = 4
for i in range(15):
ds = res_correct[0+h*i:h+h*i]
J_l = np.array(ds['Out info.J_fin'].tolist())
J_means.append(np.mean(J_l))
std_errs.append(np.std(mse_l))
my_set = set(m_sig_all)
m_sig = sorted( list(my_set) )
figure = plt.figure(figsize=(12,8))
axes = figure.add_subplot (1, 1, 1)
plt.grid(True)
plt.title('J_fin vs number of neurons', fontsize=15)
plt.xlabel('Number of neurons', fontsize=15)
plt.ylabel('J_fin', fontsize=15)
plt.scatter(m_sig_all, mse_all, label = 'J_fin for all points', marker = "D",s=40)
plt.plot(m_sig, mse_means, color='black', marker='x', linestyle='dashed', linewidth=3, markersize=16, label = 'Mean J_fin')
plt.errorbar(m_sig, mse_means, yerr=std_errs, ecolor='r', lw=2, capsize=15, mew = 3, zorder=3, label = 'Std.error of J_fin', linestyle='None')
axes.set_yscale ('log')
plt.legend(loc=1, prop={'size': 14})
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
res_n_sig5 = res_correct[res_correct['Model info.m_train'] == 16]
#m_train_all = np.array(res_n_sig5['Model info.m_train'].tolist())
#res_n_sig5 = res_correct
m_sig_all = np.array(res_n_sig5['Model info.n_sig'].tolist())
mse_all = np.array(res_n_sig5['Out info.MSE'].tolist())
mse_means = []
std_errs = []
h = 8
for i in range(13):
ds = res_n_sig5[0+h*i:8+h*i]
mse_l = np.array(ds['Out info.MSE'].tolist())
mse_means.append(np.mean(mse_l))
std_errs.append(np.std(mse_l))
my_set = set(m_sig_all)
m_sig = sorted( list(my_set) )
#print(m_sig)
figure = plt.figure(figsize=(12,8))
axes = figure.add_subplot (1, 1, 1)
plt.grid(True)
plt.title('MSE vs number of sigmoids', fontsize=15)
plt.xlabel('Number of sigmoids', fontsize=15)
plt.ylabel('MSE', fontsize=15)
plt.scatter(m_sig_all, mse_all, label = 'MSE for all points', marker = "D",s=40)
plt.plot(m_sig, mse_means, color='black', marker='x', linestyle='dashed', linewidth=3, markersize=16, label = 'Mean MSE')
plt.errorbar(m_sig, mse_means, yerr=std_errs, ecolor='r', lw=2, capsize=15, mew = 3, zorder=3, label = 'Std.error of MSE', linestyle='None')
axes.set_yscale ('log')#, nonposy='clip')
plt.legend(loc=2, prop={'size': 19})
#print(std_errs)
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]:
In [ ]: