In [38]:

    

import pandas as pd
import seaborn
import matplotlib.pyplot as plt
seaborn.set()
import numpy as np
np.set_printoptions(precision=3)
%matplotlib inline


    

In [39]:

    

import os
import glob
import re

def load_lengths(idx):
    
    df = pd.DataFrame(index=idx, columns=['duration'])
    
    for jf in idx:
        # get the basename
        match = re.match('SALAMI_(?P<tracknum>\d+).jams', jf)
        
        if not match:
            raise ValueError('Unmatched file: {}'.format(jf))
            
        # find an annotation file
        annotation = '/home/bmcfee/data/SALAMI/data_cleaned/{:s}/parsed/textfile1_functions.lab'.format(match.group('tracknum'))
        
        data = pd.read_csv(annotation, header=None, sep='\t')
        df['duration'][jf] = data[1].max()
    
    return df


    

In [40]:

    

import os
import glob
import re

def load_min_lengths(idx):
    
    small_df = pd.Series()
    large_df = pd.Series()
    
    for jf in idx:
        # get the basename
        match = re.match('SALAMI_(?P<tracknum>\d+).jams', jf)
        
        if not match:
            raise ValueError('Unmatched file: {}'.format(jf))
            
        # find an annotation file
        for annotator in [1, 2]:
            a1 = '/home/bmcfee/data/SALAMI/data_cleaned/{:s}/parsed/textfile{:d}_small.lab'.format(match.group('tracknum'), annotator)
            data = pd.read_csv(a1, header=None, sep='\t')
            small_df = pd.concat([small_df, data[1] - data[0]])

            a1 = '/home/bmcfee/data/SALAMI/data_cleaned/{:s}/parsed/textfile{:d}_functions.lab'.format(match.group('tracknum'), annotator)
            data = pd.read_csv(a1, header=None, sep='\t')
            large_df = pd.concat([large_df, data[1] - data[0]])

    return small_df, large_df


    

In [41]:

    

# df = pd.read_csv('/home/bmcfee/git/hier_eval/notes/searching_w.csv')
df = pd.read_csv('searching_w_nontransitive.csv')


    

In [42]:

    

df = df.drop('Unnamed: 0', axis=1)


    

In [43]:

    

df = df[np.argsort(df.columns)]


    

In [44]:

    

df = df.set_index('file')


    

In [45]:

    

df = df.sort_index()


    

In [ ]:

    

lengths = load_lengths(df.index)


    

In [10]:

    

small, func = load_min_lengths(df.index)


    

In [11]:

    

small.describe(percentiles=[0.5, 0.95])


    

    
Out[11]:





count    27716.000000
mean         7.912020
std          9.368818
min          0.001000
50%          6.664000
95%         15.686750
max        543.951000
dtype: float64

In [12]:

    

func.describe()


    

    
Out[12]:





count    8347.000000
mean       26.271659
std        35.003563
min         0.001000
25%         9.497500
50%        18.235000
75%        30.201500
max       615.315000
dtype: float64

In [13]:

    

df = df.join(lengths)


    

In [14]:

    

for k in ['0.5', '03', '05', '10', '15', '20', '25', '30', '60']:
    plt.figure()
    seaborn.jointplot(df['To_{}'.format(k)], df['Tu_{}'.format(k)], kind='kde')
    plt.title('w={}'.format(k))
    plt.tight_layout()


    

    






<matplotlib.figure.Figure at 0x7f2f1434bb10>






    













    






<matplotlib.figure.Figure at 0x7f2ee72dd7d0>






    













    






<matplotlib.figure.Figure at 0x7f2ee6fce710>






    













    






<matplotlib.figure.Figure at 0x7f2ee73f9510>






    













    






<matplotlib.figure.Figure at 0x7f2ee69286d0>






    













    






<matplotlib.figure.Figure at 0x7f2ee6741390>






    













    






<matplotlib.figure.Figure at 0x7f2ee6348b90>






    













    






<matplotlib.figure.Figure at 0x7f2ee6074bd0>






    













    






<matplotlib.figure.Figure at 0x7f2ee5d2abd0>






    

In [15]:

    

for k in ['0.5', '03', '05', '10', '15', '20', '25', '30', '60']:
    plt.figure()
    seaborn.jointplot(df['duration'], df['Tf_{}'.format(k)], kind='reg')
    plt.title('w={}'.format(k))
    plt.tight_layout()


    

    






<matplotlib.figure.Figure at 0x7f2ee6043ed0>






    













    






<matplotlib.figure.Figure at 0x7f2ee5746050>






    













    






<matplotlib.figure.Figure at 0x7f2ee53dfcd0>






    













    






<matplotlib.figure.Figure at 0x7f2ee50e3510>






    













    






<matplotlib.figure.Figure at 0x7f2ee4d91610>






    













    






<matplotlib.figure.Figure at 0x7f2ee5c8e950>






    













    






<matplotlib.figure.Figure at 0x7f2ee4d99850>






    













    






<matplotlib.figure.Figure at 0x7f2ee6172890>






    













    






<matplotlib.figure.Figure at 0x7f2ee64d61d0>






    

In [46]:

    

t_f = df[df.columns[:14]]


    

In [47]:

    

t_o = df[df.columns[14:28]]
t_u = df[df.columns[28:42]]


    

In [48]:

    

cols = t_f.columns


    

In [49]:

    

t_f.columns = [x.replace('Tf_', '') for x in cols]


    

In [50]:

    

seaborn.set(style='ticks')


    

In [51]:

    

seaborn.set_style('darkgrid')
plt.figure(figsize=(6,3))
seaborn.boxplot(t_f, color=seaborn.color_palette('Blues')[3])
plt.xlabel('$w$ (seconds)')
plt.ylabel('$\mathcal{T}_F$')
plt.savefig('../figs/tfw.pdf', bbox_inches='tight', pad_inches=0.05)


    

In [233]:

    

plt.figure(figsize=(6,4))
seaborn.boxplot(t_o, color=seaborn.color_palette('Blues')[3])
plt.xlabel('$w$')
plt.ylabel('$\mathcal{T}_O$')
#plt.savefig('../figs/tfw.pdf')


    

    
Out[233]:





<matplotlib.text.Text at 0x7fb02041f810>






    

In [234]:

    

plt.figure(figsize=(6,4))
seaborn.boxplot(t_u, color=seaborn.color_palette('Blues')[3])
plt.xlabel('$w$')
plt.ylabel('$\mathcal{T}_U$')
#plt.savefig('../figs/tfw.pdf')


    

    
Out[234]:





<matplotlib.text.Text at 0x7fb0214162d0>