Show topic distributions

Plot topic figures and show them.

Start with the necessary imports

from os.path import join

import numpy as np
import pandas as pd
from nilearn import plotting
from nilearn.masking import unmask

from gclda.model import Model
from gclda.utils import get_resource_path

Load model

model_file = join(get_resource_path(), 'models/Neurosynth2015Filtered2',
                  'model_200topics_2015Filtered2_10000iters.pklz')
model = Model.load(model_file)

Get spatial probability

p_voxel_g_topic, _ = model.get_spatial_probs()

Show topic 10

topic_no = 10
img = unmask(p_voxel_g_topic[:, topic_no], model.dataset.mask_img)

# Get strings giving top K words and probs for the current topic
wprobs = model.n_word_tokens_word_by_topic[:, topic_no] + model.beta
wprobs = wprobs / np.sum(wprobs)
word_probs = zip(*[model.dataset.word_labels, wprobs])
df = pd.DataFrame(columns=['term', 'probability'], data=word_probs)
df = df.sort_values(by='probability', ascending=False).reset_index(drop=True)
print df.head(12)

fig = plotting.plot_stat_map(img, display_mode='z')

Out:

term  probability
0   cognitive_control     0.119603
1             demands     0.071321
2           executive     0.062462
3           difficult     0.033669
4                easy     0.024810
5   executive_control     0.023924
6           requiring     0.019938
7           selection     0.018609
8              number     0.016837
9                hard     0.015951
10             manner     0.015508
11           matching     0.015065

Show topic 59

topic_no = 58
img = unmask(p_voxel_g_topic[:, topic_no], model.dataset.mask_img)

# Get strings giving top K words and probs for the current topic
wprobs = model.n_word_tokens_word_by_topic[:, topic_no] + model.beta
wprobs = wprobs / np.sum(wprobs)
word_probs = zip(*[model.dataset.word_labels, wprobs])
df = pd.DataFrame(columns=['term', 'probability'], data=word_probs)
df = df.sort_values(by='probability', ascending=False).reset_index(drop=True)
print df.head(12)

fig = plotting.plot_stat_map(img, display_mode='z')

Out:

term  probability
0    somatosensory     0.135916
1          tactile     0.108811
2            touch     0.088289
3      stimulation     0.048406
4          sensory     0.039113
5   discrimination     0.027496
6             body     0.022850
7     vibrotactile     0.014331
8          touched     0.013556
9           finger     0.013169
10            skin     0.012395
11       roughness     0.012008

Show topic 150

topic_no = 149
img = unmask(p_voxel_g_topic[:, topic_no], model.dataset.mask_img)

# Get strings giving top K words and probs for the current topic
wprobs = model.n_word_tokens_word_by_topic[:, topic_no] + model.beta
wprobs = wprobs / np.sum(wprobs)
word_probs = zip(*[model.dataset.word_labels, wprobs])
df = pd.DataFrame(columns=['term', 'probability'], data=word_probs)
df = df.sort_values(by='probability', ascending=False).reset_index(drop=True)
print df.head(12)

fig = plotting.plot_stat_map(img, display_mode='z')

Out:

term  probability
0    emotional     0.120525
1      emotion     0.064047
2        faces     0.055130
3    affective     0.027836
4         fear     0.025134
5      fearful     0.024594
6     pictures     0.015136
7   regulation     0.014055
8       affect     0.013784
9        angry     0.012704
10      threat     0.012704
11     ratings     0.012704

Total running time of the script: ( 0 minutes 41.400 seconds)