-
Notifications
You must be signed in to change notification settings - Fork 0
/
captum_viz_json3.py
175 lines (147 loc) · 6.71 KB
/
captum_viz_json3.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
import json
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from wordcloud import WordCloud
import umap
def load_json_data(file_path):
with open(file_path, 'r') as f:
return json.load(f)
def create_dataframe(data):
df = pd.DataFrame(data)
df['attributions'] = df['attributions'].apply(np.array)
return df
def pad_sequence(seq, max_length):
return np.pad(seq, (0, max_length - len(seq)), 'constant')
def pad_attributions(df):
max_length = df['attributions'].apply(len).max()
df['padded_attributions'] = df['attributions'].apply(lambda x: pad_sequence(x, max_length))
return df
def plot_attribution_heatmap(df, n_samples=100, figsize=(20, 10), group=''):
plt.figure(figsize=figsize)
sample_df = df.sample(n_samples)
padded_df = pad_attributions(sample_df)
heatmap_data = np.vstack(padded_df['padded_attributions'])
sns.heatmap(heatmap_data, cmap='coolwarm', center=0)
plt.title(f'Attribution Heatmap - {group} (Sample of {n_samples} sequences)')
plt.xlabel('Token Position')
plt.ylabel('Sequence')
plt.savefig(f'attribution_heatmap_{group.lower()}.svg', format='svg')
plt.close()
def plot_average_attribution(df, top_n=20, figsize=(12, 6), group=''):
padded_df = pad_attributions(df)
avg_attributions = np.mean(np.vstack(padded_df['padded_attributions']), axis=0)
top_indices = np.argsort(np.abs(avg_attributions))[-top_n:]
plt.figure(figsize=figsize)
sns.barplot(x=avg_attributions[top_indices], y=np.arange(top_n))
plt.title(f'Top {top_n} Average Attributions - {group}')
plt.xlabel('Average Attribution')
plt.ylabel('Token Position')
plt.savefig(f'average_attribution_{group.lower()}.svg', format='svg')
plt.close()
def plot_probability_distribution(df, figsize=(10, 6), group=''):
plt.figure(figsize=figsize)
sns.histplot(df['positive_probability'], kde=True)
plt.title(f'Distribution of Positive Probabilities - {group}')
plt.xlabel('Positive Probability')
plt.ylabel('Count')
plt.savefig(f'probability_distribution_{group.lower()}.svg', format='svg')
plt.close()
def plot_token_importance(df, top_n=20, figsize=(12, 8), group=''):
token_importance = {}
for _, row in df.iterrows():
for token, attr in zip(row['tokens'], row['attributions']):
if token in token_importance:
token_importance[token].append(abs(attr))
else:
token_importance[token] = [abs(attr)]
avg_importance = {k: np.mean(v) for k, v in token_importance.items()}
top_tokens = sorted(avg_importance.items(), key=lambda x: x[1], reverse=True)[:top_n]
plt.figure(figsize=figsize)
sns.barplot(x=[t[1] for t in top_tokens], y=[t[0] for t in top_tokens])
plt.title(f'Top {top_n} Important Tokens - {group}')
plt.xlabel('Average Absolute Attribution')
plt.ylabel('Token')
plt.savefig(f'token_importance_{group.lower()}.svg', format='svg')
plt.close()
def plot_attribution_clusters(df, n_clusters=5, figsize=(12, 8), group=''):
padded_df = pad_attributions(df)
X = np.vstack(padded_df['padded_attributions'])
kmeans = KMeans(n_clusters=n_clusters)
clusters = kmeans.fit_predict(X)
pca = PCA(n_components=2)
X_pca = pca.fit_transform(X)
plt.figure(figsize=figsize)
scatter = plt.scatter(X_pca[:, 0], X_pca[:, 1], c=clusters, cmap='viridis')
plt.colorbar(scatter)
plt.title(f'Attribution Clusters - {group}')
plt.xlabel('PCA Component 1')
plt.ylabel('PCA Component 2')
plt.savefig(f'attribution_clusters_{group.lower()}.svg', format='svg')
plt.close()
def generate_wordcloud(df, figsize=(12, 8), group=''):
text = ' '.join([' '.join(tokens) for tokens in df['tokens']])
wordcloud = WordCloud(width=800, height=400, background_color='white').generate(text)
plt.figure(figsize=figsize)
plt.imshow(wordcloud, interpolation='bilinear')
plt.axis('off')
plt.title(f'Word Cloud of Tokens - {group}')
plt.savefig(f'token_wordcloud_{group.lower()}.svg', format='svg')
plt.close()
def plot_sequence_length_distribution(df, figsize=(10, 6), group=''):
sequence_lengths = df['tokens'].apply(len)
plt.figure(figsize=figsize)
sns.histplot(sequence_lengths, kde=True)
plt.title(f'Distribution of Sequence Lengths - {group}')
plt.xlabel('Sequence Length')
plt.ylabel('Count')
plt.savefig(f'sequence_length_distribution_{group.lower()}.svg', format='svg')
plt.close()
def generate_summary_stats(df, group=''):
summary = {
'total_sequences': len(df),
'avg_positive_probability': df['positive_probability'].mean(),
'median_positive_probability': df['positive_probability'].median(),
'avg_sequence_length': df['tokens'].apply(len).mean(),
'median_sequence_length': df['tokens'].apply(len).median(),
}
with open(f'summary_stats_{group.lower()}.json', 'w') as f:
json.dump(summary, f, indent=2)
def plot_umap(df, n_neighbors=15, min_dist=0.1, figsize=(12, 8), group=''):
padded_df = pad_attributions(df)
X = np.vstack(padded_df['padded_attributions'])
reducer = umap.UMAP(n_neighbors=n_neighbors, min_dist=min_dist, random_state=42)
embedding = reducer.fit_transform(X)
plt.figure(figsize=figsize)
scatter = plt.scatter(embedding[:, 0], embedding[:, 1], c=df['positive_probability'], cmap='viridis')
plt.colorbar(scatter, label='Positive Probability')
plt.title(f'UMAP Projection of Attributions - {group}')
plt.xlabel('UMAP Dimension 1')
plt.ylabel('UMAP Dimension 2')
plt.savefig(f'umap_projection_{group.lower()}.svg', format='svg')
plt.close()
def generate_visualizations(df, group):
plot_attribution_heatmap(df, group=group)
plot_average_attribution(df, group=group)
plot_probability_distribution(df, group=group)
plot_token_importance(df, group=group)
plot_attribution_clusters(df, group=group)
generate_wordcloud(df, group=group)
plot_sequence_length_distribution(df, group=group)
generate_summary_stats(df, group=group)
plot_umap(df, group=group) # Add this line to include UMAP analysis
def main():
# Load and process algal data
algal_data = load_json_data('algal_captum_output.json')
algal_df = create_dataframe(algal_data)
# Load and process bacteria data
bacteria_data = load_json_data('bacteria_captum_output.json')
bacteria_df = create_dataframe(bacteria_data)
# Generate visualizations for algal data
generate_visualizations(algal_df, 'Algal')
# Generate visualizations for bacteria data
generate_visualizations(bacteria_df, 'Bacterial')
if __name__ == "__main__":