In [1]:

    

import numpy as np
import quail
from quail import Fingerprint, OptimalPresenter
import seaborn as sns
from scipy.spatial.distance import cdist
import pandas as pd


    

In [2]:

    

# nperms=100
# dist_stick = []
# dist_beststick = []
# dist_perm = []
# dist_permde = []
# dist_rand = []
# dist_bestchoice = []
    
# for iperm in range(nperms):
    
#     presenter.strategy = 'stabilize'
    
# #     # random
#     shuffled_egg = presenter.order(egg, strategy='random')
#     weights_rand = quail.analyze(shuffled_egg, analysis='fingerprint')
#     dist_rand.append(cdist(weights_rand, weights)[0])
    
#     # reorder next list using stick breaking
#     resorted_egg = presenter.order(shuffled_egg, method='stick')
#     weights_stick = quail.analyze(resorted_egg, analysis='fingerprint')
#     dist_stick.append(cdist(weights_stick, weights)[0])

#     # reorder next list using stick breaking
#     resorted_egg = presenter.order(shuffled_egg, method='best_stick', nperms=10000)
#     weights_stick = quail.analyze(resorted_egg, analysis='fingerprint')
#     dist_beststick.append(cdist(weights_stick, weights)[0])
    
#     # reorder next list using permutation
#     resorted_egg = presenter.order(shuffled_egg, method='permute', nperms=10000)
#     weights_perm = quail.analyze(resorted_egg, analysis='fingerprint')
#     dist_perm.append(cdist(weights_perm, weights))
    
#     presenter.strategy = 'destabilize'
    
#     # reorder next list using permutation
#     resorted_egg = presenter.order(shuffled_egg, method='permute', nperms=10000)
#     weights_permde = quail.analyze(resorted_egg, analysis='fingerprint')
#     dist_permde.append(cdist(weights_permde, weights))
    
#     # reorder next list using permutation
# #     resorted_egg = presenter.order(shuffled_egg, method='best_choice', nperms=100)
# #     weights_choice = quail.analyze(resorted_egg, analysis='fingerprint')
# #     dist_bestchoice.append(cdist(weights_choice, weights)[0])
    
#     if iperm % 10 == 0:
#         print(iperm)


    

In [3]:

    

# sns.distplot(dist_stick, label='stick')
# sns.distplot(dist_beststick, label='best stick')
# # sns.distplot(dist_bestchoice, label='best choice')
# sns.distplot(dist_perm, label='permute:stabilize')
# sns.distplot(dist_permde, label='permute:destabilize')
# sns.distplot(dist_rand, label='random')
# sns.plt.legend()
# sns.plt.show()


    

In [4]:

    

def loadstim():
    """Loads in the stimuli from disk and creates an egg"""

    def shuffle_stim(wordpool, nlists=16, nwords=16):
        """Randomizes the stimuli"""
        np.random.seed()
        return pd.concat([wp[wp['GROUP']==choice].sample(nwords, replace=False).reset_index(drop=True) for choice in np.random.choice(wp['GROUP'].unique(), nlists, replace=False)])

    def add_features(wp):
        """Adds features to the stimuli"""

        # first letter
        wp['FIRST LETTER'] = wp['WORD'].apply(lambda x: x[0])

        # word length
        wp['WORD LENGTH'] = wp['WORD'].apply(lambda x: len(x))

        # color
        wp['COLOR'] = [[int(np.random.rand() * 255) for i in range(3)] for i in range(wp.shape[0])]

        # location
        wp['LOCATION'] = [[np.random.rand() * 85, np.random.rand() * (100 - wp['WORD LENGTH'].iloc[i] * 3)] for i in range(wp.shape[0])]

        return wp

    def df2egg(df):
        """Converts a dataframe to an egg"""
        group = df['GROUP'].unique()
        pres = [df[df['GROUP']==i]['WORD'].values.tolist() for i in group]
        rec = [[None] for i in group]
        features = [format_features(df[df['GROUP']==i]) for i in group]
        return quail.Egg(pres=[pres], rec=[rec], features=[features])

    def format_features(wp):
        """Formats the features"""
        trials=[]
        for idx, row in wp.iterrows():
            trial = {
                'length' : row['WORD LENGTH'],
                'size' : row['SIZE'],
                'category' : row['CATEGORY'],
                'pos' : row['LOCATION'],
                'color' : row['COLOR'],
                'first_letter' : row['WORD'][0]
            }
            trials.append(trial)
        return trials

    # load wordpool
    wp = pd.read_csv('/Users/andyheusser/Documents/github/adaptiveFR/exp/static/files/cut_wordpool.csv')

    # shuffle the lists and stim
    shuffled_stim = shuffle_stim(wp)

    # add features
    stim_features = add_features(shuffled_stim)

    # go from df to egg
    egg = df2egg(stim_features)

    return egg

# create the stim list
egg = loadstim()
egg.dist_funcs


    

    
Out[4]:





{'category': <function quail.helpers.<lambda>>,
 'color': <function quail.helpers.<lambda>>,
 'first_letter': <function quail.helpers.<lambda>>,
 'length': <function quail.helpers.<lambda>>,
 'pos': <function quail.helpers.<lambda>>,
 'size': <function quail.helpers.<lambda>>}

In [14]:

    

%%timeit

egg1 = egg.crack(lists=[0], subjects=[0])
egg2 = egg.crack(lists=[1], subjects=[0])

# initialize fingerprint
fingerprint = Fingerprint(init=egg1)

# initialize presenter
params = {
    'fingerprint' : fingerprint
}
presenter = OptimalPresenter(params=params, strategy='stabilize')

# update the fingerprint
fingerprint.update(egg1, permute=True, parallel=True)

# reshape weights
weights = fingerprint.state.reshape(1,6)

# reorder next list using permutation
resorted_egg = presenter.order(egg2, method='permute', nperms=2500)


    

    




1 loop, best of 3: 20.4 s per loop

In [4]:

    

nperms=100
dist_stick = []
dist_beststick = []
dist_perm = []
dist_permde = []
dist_rand = []
dist_bestchoice = []
    
for iperm in range(nperms):
    
    # generate a list from the word pool
    lst1 = generate_lst(wp)
    
    # simulate an egg
    egg1 = simulate_egg(lst1)
    
    # initialize fingerprint
    fingerprint = Fingerprint(init=egg1)

    # initialize presenter
    params = {
        'fingerprint' : fingerprint
    }
    presenter = OptimalPresenter(params=params, strategy='stabilize')

    # update the fingerprint
    fingerprint.update(egg1, permute=True, parallel=True)
    
    # reshape weights
    weights = fingerprint.state.reshape(1,7)
    
    # generate another list from the word pool
    lst2 = generate_lst(wp)
    
    # simulate egg 2
    egg2 = simulate_egg(lst2)

    weights_rand = quail.analyze(egg2, analysis='fingerprint', permute=True, parallel=True)
    dist_rand.append(cdist(weights_rand, weights, 'correlation')[0])
    
    # reorder next list using stick breaking
    resorted_egg = presenter.order(egg2, method='stick')
    weights_stick = quail.analyze(resorted_egg, analysis='fingerprint', permute=True, parallel=True)
    dist_stick.append(cdist(weights_stick, weights, 'correlation')[0])

    # reorder next list using stick breaking
    resorted_egg = presenter.order(egg2, method='best_stick', nperms=nperms)
    weights_stick = quail.analyze(resorted_egg, analysis='fingerprint', permute=True, parallel=True)
    dist_beststick.append(cdist(weights_stick, weights, 'correlation')[0])
    
    # reorder next list using permutation
    resorted_egg = presenter.order(egg2, method='permute', nperms=nperms)
    weights_perm = quail.analyze(resorted_egg, analysis='fingerprint', permute=True, parallel=True)
    dist_perm.append(cdist(weights_perm, weights, 'correlation'))
    
    presenter.strategy = 'destabilize'
    
    # reorder next list using permutation
    resorted_egg = presenter.order(egg2, method='permute', nperms=nperms)
    weights_permde = quail.analyze(resorted_egg, analysis='fingerprint', permute=True, parallel=True)
    dist_permde.append(cdist(weights_permde, weights, 'correlation'))

    if iperm % 10 == 0:
        print(iperm)


    

    




/Users/andyheusser/Documents/github/quail/quail/analysis.py:105: UserWarning: Pandas doesn't allow columns to be created via a new attribute name - see https://pandas.pydata.org/pandas-docs/stable/indexing.html#attribute-access
  'list_length' : data.list_length






    




0
10






    




Process PoolWorker-4:
Process PoolWorker-2:
Process PoolWorker-3:
Process PoolWorker-1:
Traceback (most recent call last):
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
Traceback (most recent call last):
Traceback (most recent call last):
Traceback (most recent call last):
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
    self.run()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
    self.run()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
    self.run()
    self.run()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
    self._target(*self._args, **self._kwargs)
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
    self._target(*self._args, **self._kwargs)
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
    task = get()
    task = get()
    task = get()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
    task = get()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
    racquire()
    racquire()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 381, in get
KeyboardInterrupt
    racquire()
    return recv()
KeyboardInterrupt
KeyboardInterrupt
KeyboardInterrupt






    




---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-4-d02dbef354d6> in <module>()
     57 
     58     # reorder next list using permutation
---> 59     resorted_egg = presenter.order(egg2, method='permute', nperms=nperms)
     60     weights_permde = quail.analyze(resorted_egg, analysis='fingerprint', permute=True, parallel=True)
     61     dist_permde.append(cdist(weights_permde, weights, 'correlation'))

/Users/andyheusser/Documents/github/quail/quail/fingerprint.py in order(self, egg, method, nperms, strategy, distfun)
    318             return shuffle_egg(egg)
    319         elif method=='permute':
--> 320             return order_perm(self, egg, dist_dict, strategy, nperms, distfun) #
    321         elif method=='stick':
    322             return order_stick(self, egg, dist_dict, strategy) #

/Users/andyheusser/Documents/github/quail/quail/fingerprint.py in order_perm(self, egg, dist_dict, strategy, nperm, distperm)
    242             pres_len = len(pres)
    243 
--> 244             results = Parallel(n_jobs=multiprocessing.cpu_count())(
    245             delayed(rand_perm)(pres, features, dist_dict, dist_funcs) for i in range(nperms))
    246 

/Applications/Canopy.app/appdata/canopy-1.7.4.3348.macosx-x86_64/Canopy.app/Contents/lib/python2.7/multiprocessing/__init__.pyc in cpu_count()
    122         try:
    123             with os.popen(comm) as p:
--> 124                 num = int(p.read())
    125         except ValueError:
    126             num = 0

KeyboardInterrupt: 





    




Process PoolWorker-7:
Process PoolWorker-5:
Process PoolWorker-8:
Process PoolWorker-6:
Traceback (most recent call last):
Traceback (most recent call last):
Traceback (most recent call last):
Traceback (most recent call last):
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
    self.run()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 258, in _bootstrap
    self.run()
    self.run()
    self.run()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
    self._target(*self._args, **self._kwargs)
    self._target(*self._args, **self._kwargs)
    self._target(*self._args, **self._kwargs)
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
    task = get()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/pool.py", line 102, in worker
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 381, in get
    return recv()
    task = get()
    task = get()
KeyboardInterrupt
    task = get()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
    racquire()
  File "/Users/andyheusser/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/multiprocess/queues.py", line 379, in get
KeyboardInterrupt
    racquire()
    racquire()
KeyboardInterrupt
KeyboardInterrupt

In [67]:

    

%timeit
shuffled_egg = presenter.order(egg1, strategy='random')
resorted_egg = presenter.order(shuffled_egg, method='permute', nperms=2500)
weights_perm = quail.analyze(resorted_egg, analysis='fingerprint')


    

    




---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-67-8974ec0814c0> in <module>()
      1 get_ipython().magic(u'timeit')
----> 2 shuffled_egg = presenter.order(egg1, strategy='random')
      3 resorted_egg = presenter.order(shuffled_egg, method='permute', nperms=2500)
      4 weights_perm = quail.analyze(resorted_egg, analysis='fingerprint')

/Users/andyheusser/Documents/github/quail/quail/fingerprint.py in order(self, egg, method, nperms, strategy, distfun)
    322             strategy = self.strategy
    323 
--> 324         dist_dict = compute_distances_dict(egg)
    325 
    326         if (strategy=='random') or (method=='random'):

/Users/andyheusser/Documents/github/quail/quail/fingerprint.py in compute_distances_dict(egg)
    651             for feature in dist_funcs:
    652 
--> 653                 distances[item1][item2][feature] = dist_funcs[feature](features_list[idx1][feature],features_list[idx2][feature])
    654 
    655     return distances

KeyError: 'CATEGORY'

In [6]:

    

%matplotlib inline
import matplotlib.pyplot as plt
sns.distplot(1-np.array(dist_stick), label='stick')
sns.distplot(1-np.array(dist_beststick), label='best stick')
# sns.distplot(dist_bestchoice, label='best choice')
sns.distplot(1-np.array(dist_perm), label='permute:stabilize')
sns.distplot(1-np.array(dist_permde), label='permute:destabilize')
sns.distplot(1-np.array(dist_rand), label='random')
plt.legend()


    

    
Out[6]:





<matplotlib.legend.Legend at 0x10af4ba10>






    

In [25]:

    

def shuffle_stim(wordpool, nlists=16, nwords=16):

    return pd.concat([wp[wp['GROUP']==choice].sample(nwords).reset_index(drop=True) for choice in np.random.choice(wp['GROUP'].unique(), nlists, replace=False)])

def add_features(wp):
    
    # first letter
    wp['FIRST LETTER'] = wp['WORD'].apply(lambda x: x[0])

    # word length
    wp['WORD LENGTH'] = wp['WORD'].apply(lambda x: len(x))

    # color
    wp['COLOR'] = [[int(np.random.rand() * 255) for i in range(3)] for i in range(wp.shape[0])]

    # location
    wp['LOCATION'] = [[np.random.rand() * 85, np.random.rand() * (100 - wp['WORD LENGTH'].iloc[i] * 3)] for i in range(wp.shape[0])]

    return wp

def df2egg(df):
    group = df['GROUP'].unique()
    pres = [df[df['GROUP']==i]['WORD'].values.tolist() for i in group]
    rec = [[None] for i in group]
    features = [format_features(df[df['GROUP']==i]) for i in group]
    return quail.Egg(pres=[pres], rec=[rec], features=[features])
    
def format_features(wp):
    trials=[]
    for idx, row in wp.iterrows():
        trial = {
            'length' : row['WORD LENGTH'],
            'size' : row['SIZE'],
            'category' : row['CATEGORY'],
            'pos' : row['LOCATION'],
            'color' : row['COLOR'],
            'first letter' : row['WORD'][0]
        }
        trials.append(trial)
    return trials

# load wordpool
wp = pd.read_csv('/Users/andyheusser/Documents/github/quail/quail/data/cut_wordpool.csv')

# shuffle the lists and stim
shuffled_stim = shuffle_stim(wp)

# add features
stim_features = add_features(shuffled_stim)

egg = df2egg(stim_features)


    

In [26]:

    

egg.dist_funcs


    

    
Out[26]:





{'CATEGORY': <function quail.helpers.<lambda>>,
 'COLOR': <function quail.helpers.<lambda>>,
 'FIRST LETTER': <function quail.helpers.<lambda>>,
 'GROUP': <function quail.helpers.<lambda>>,
 'LOCATION': <function quail.helpers.<lambda>>,
 'SIZE': <function quail.helpers.<lambda>>,
 'WORD LENGTH': <function quail.helpers.<lambda>>,
 'category': <function quail.helpers.<lambda>>,
 'color': <function quail.helpers.<lambda>>,
 'first letter': <function quail.helpers.<lambda>>,
 'length': <function quail.helpers.<lambda>>,
 'pos': <function quail.helpers.<lambda>>,
 'size': <function quail.helpers.<lambda>>}

In [6]:

    

# generate a list from the word pool
lst1 = generate_lst(wp)

# simulate an egg
egg = simulate_egg(lst1)


    

In [ ]:

    

# def format_stim(wp):
#     trials=[]
#     for idx, row in wp.iterrows():
#         trial = {
#             'type' : 'p',
#             'text' : row['WORD'],
#             'length' : row['WORD LENGTH'],
#             'size' : row['SIZE'],
#             'category' : row['CATEGORY'],
#             'listid' : row['GROUP'],
#             'pos' : row['LOCATION'],
#             'rgb' : row['COLOR'],
#         }
#         trials.append(trial)
#     return trials