-
Notifications
You must be signed in to change notification settings - Fork 0
/
performance.py
385 lines (362 loc) · 13.7 KB
/
performance.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
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
from pathlib import Path
from functools import wraps
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.metrics import roc_curve, auc
from tensorflow.keras.utils import to_categorical
import pycm
import wandb
"""
Metrics
=======
"""
CLASS_METRICS = [
"TP", "FP", "TN", "FN",
"PPV", # Precision
"TPR", # Recall
"ACC", # Accuracy
"F1" # F1 score
]
OVERALL_METRICS = [
"ACC Macro", # - Accuracy Macro
"ERR Macro", # - Error Macro
"PPV Macro", # - Precision Macro
"TNR Macro", # - Specificity Macro
"TPR Macro", # - Recall Macro
"F1 Macro", # - F1-Score Macro
"Kappa" # - Kappa
]
def _filter_and_validate_stats(stats, metrics):
filtered_stats = {}
for m in metrics:
if isinstance(stats[m], dict):
filtered_stats[m] = {
k:v if v != "None" else 0.0 for k,v in stats[m].items()
}
else:
filtered_stats[m] = stats[m] if stats[m] != "None" else 0.0
return filtered_stats
def classification_metrics_by_class(y_true, y_pred, labels=None, flatten=True, metrics=CLASS_METRICS):
"""get classification metrics by class for binary and multi-class.
metrics:
"TP", "FP", "TN", "FN",
"PPV", # Precision
"TPR", # Recall
"ACC", # Accuracy
"F1" # F1 score
"""
cm = pycm.ConfusionMatrix(actual_vector=y_true, predict_vector=y_pred)
if labels is not None:
#if isinstance(labels, list) or isinstance(labels, np.ndarray):
if not isinstance(labels, dict):
labels = {i:str(l) for i, l in enumerate(labels)}
cm.relabel(mapping=labels)
stats = _filter_and_validate_stats(cm.class_stat, metrics)
if flatten:
flatten_stats = {}
for metric, by_class in stats.items():
flatten_stats.update({f"{metric} - {c}": m for c, m in by_class.items()})
stats = flatten_stats
return stats
def overall_classification_metrics(y_true, y_pred, labels=None, metrics=OVERALL_METRICS):
"""get overall classification metrics for multi-class.
metrics:
"ACC Macro", # - Accuracy Macro
"ERR Macro", # - Error Macro
"PPV Macro", # - Precision Macro
"TNR Macro", # - Specificity Macro
"TPR Macro", # - Recall Macro
"F1 Macro", # - F1-Score Macro
"Kappa" # - Kappa
"""
cm = pycm.ConfusionMatrix(actual_vector=y_true, predict_vector=y_pred)
if labels is not None:
#if isinstance(labels, list):
if not isinstance(labels, dict):
labels = {i:str(l) for i, l in enumerate(labels)}
cm.relabel(mapping=labels)
stats = cm.overall_stat
#add error rate
stats["ERR Macro"] = sum(cm.ERR.values())/len(cm.ERR)
stats = _filter_and_validate_stats(stats, metrics)
return stats
def get_metrics(y_true, y_prob, metrics, labels=None, num_classes=None, run_id=None):
num_classes = num_classes or y_prob.shape[-1]
y_pred = y_prob.argmax(axis=-1) if num_classes > 2 else np.round(y_prob).astype(int)
all_metrics = {}
tables_metrics = {}
base_columns = [] if run_id is None else ['Fold']
base_data = [] if run_id is None else [run_id]
if "overall_classification_metrics" in metrics:
overall_metrics = overall_classification_metrics(y_true, y_pred, labels=labels)
all_metrics.update(overall_metrics)
tables_metrics["overall_metrics_table"] = wandb.Table(
columns=base_columns + list(overall_metrics.keys()),
data=[base_data + list(overall_metrics.values())]
)
if "classification_metrics_by_class" in metrics:
metrics_by_class = classification_metrics_by_class(y_true, y_pred, labels=labels)
all_metrics.update(metrics_by_class)
tables_metrics["by_class_metrics_table"] = wandb.Table(
columns=base_columns + list(metrics_by_class.keys()),
data=[base_data + list(metrics_by_class.values())]
)
return all_metrics, tables_metrics
"""
Plots
=====
"""
def save_or_show_plot(plot_name):
"""
Add save/show
"""
def real_decorator(plot_func):
@wraps(plot_func)
def wrapper(*args, **kwargs):
save_to = kwargs.get("save_to", None)
model_name=kwargs.get("model_name", None)
fig = plot_func(*args, **kwargs)
if save_to is not None:
save_to = Path(save_to)
fig.savefig(save_to / f'{model_name}_{plot_name}.png')
show = kwargs.get("show", False)
if show:
plt.show()
plt.close()
return None
else:
return fig
return wrapper
return real_decorator
def confusion_matrix(y_true, y_pred, labels=None):
cm = pycm.ConfusionMatrix(actual_vector=y_true, predict_vector=y_pred)
if labels:
if isinstance(labels, list):
labels = {i:str(l) for i, l in enumerate(labels)}
cm.relabel(mapping=labels)
return cm
@save_or_show_plot("Confussion Matrix")
def plot_confusion_matrix(cm, model_name=None, labels=None, cmap=None, normalize=True, *args, **kwargs):
"""
given a pycm confusion matrix (cm), make a nice plot
Arguments
---------
cm: confusion matrix from sklearn.metrics.confusion_matrix
labels: given classification classes such as [0, 1, 2]
the class names, for example: ['high', 'medium', 'low']
title: the text to display at the top of the matrix
cmap: the gradient of the values displayed from matplotlib.pyplot.cm
see http://matplotlib.org/examples/color/colormaps_reference.html
plt.get_cmap('jet') or plt.cm.Blues
normalize: If False, plot the raw numbers
If True, plot the proportions
Return
------
figure or None
Usage
-----
plot_confusion_matrix(cm = cm, # confusion matrix created by pycm
normalize = True, # show proportions
labels = y_labels_vals, # list of names of the classes
title = best_estimator_name) # title of graph
Citiation
---------
http://scikit-learn.org/stable/auto_examples/model_selection/plot_confusion_matrix.html
"""
accuracy = cm.ACC_Macro
kappa = cm.Kappa
labels = labels or cm.classes
cm = cm.to_array()
if normalize:
group_counts = [f"{value}"
for value in cm.flatten()]
group_percentages = [f"{100*value:.1f}"
for value in cm.flatten()/np.sum(cm)]
annot = [f"{v1}\n{v2}%"
for v1, v2 in zip(group_counts, group_percentages)]
annot = np.array(annot).reshape(cm.shape)
fmt =''
else:
fmt ='d'
annot=True
# Heatmap
fig = plt.figure(figsize=(5, 4))
cmap = cmap or plt.get_cmap('Blues')
sns.heatmap(cm, annot=annot, cmap=cmap, fmt=fmt,
xticklabels=labels, yticklabels=labels)
# Style
if model_name is not None:
title = f"{model_name}\nConfusion Matrix"
else:
title = f"Confusion Matrix"
plt.title(title)
plt.ylabel('True label')
plt.xlabel(f'Predicted label\n\naccuracy={accuracy:0.4f}; Kappa={kappa:0.4f}')
plt.grid(False)
plt.tight_layout()
return fig
def roc_curve_by_class(y_true, y_prob, num_classes=None):
#m_samples, n_classes = y_prob.shape
num_classes = num_classes or (y_prob.shape[-1] if y_prob.ndim > 1 else 2)
y_true = to_categorical(y_true, num_classes=num_classes)
fpr = dict()
tpr = dict()
roc_auc = dict()
for i in range(num_classes):
if num_classes == 2:
fpr[i], tpr[i], _ = roc_curve((1 - y_true[:]), (1 - y_prob[:])) if i == 0 else roc_curve(y_true[:], y_prob[:])
else:
fpr[i], tpr[i], _ = roc_curve(y_true[:, i], y_prob[:, i])
roc_auc[i] = auc(fpr[i], tpr[i])
return fpr, tpr, roc_auc
@save_or_show_plot("ROC Curve")
def plot_roc_curve_by_class(fpr, tpr, roc_auc, model_name=None, labels=None):
"""
given a sklearn roc curves, make a nice plot
Arguments
---------
fpr, tpr, roc_auc :
labels: given classification classes such as [0, 1, 2]
the class names, for example: ['high', 'medium', 'low']
title: the text to display at the top of the matrix
Return
------
figure or None
Usage
-----
display_multiclass_roc_curve(...)
Citiation
---------
https://stackoverflow.com/questions/50941223/plotting-roc-curve-with-multiple-classes
"""
n_classes = len(fpr)
labels = labels or range(n_classes)
# ROC Curves
fig = plt.figure(figsize=(5, 4))
for i in range(n_classes):
plt.plot( fpr[i], tpr[i],
label=f'ROC curve of class {labels[i]} (AUC = {roc_auc[i]:0.2f})')
# Style
if model_name is not None:
title = f"{model_name}\nReceiver Operating Characteristic"
else:
title = f"Receiver Operating Characteristic"
plt.plot([0, 1], [0, 1], 'k--')
plt.title(title)
plt.xlim([-0.05, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.legend(loc="lower right")
plt.tight_layout()
return fig
def get_plots(X_true, y_true, y_prob, plots, labels=None, num_classes=None, wandb_plot=False):
y_pred = y_prob.argmax(axis=-1) if num_classes > 2 else np.round(y_prob).astype(int)
all_plots = {}
if 'cm' in plots:
if wandb_plot:
raise NotImplementedError
else:
cm = confusion_matrix(y_true, y_pred, labels=labels)
fig_cm = plot_confusion_matrix(cm, labels=labels)
all_plots['cm'] = wandb.Image(fig_cm)
plt.close(fig_cm)
if 'pr' in plots:
if wandb_plot:
#all_plots['pr'] = wandb.plots.precision_recall(y_true, y_prob, labels)
raise NotImplementedError
else:
fpr, tpr, roc_auc = roc_curve_by_class(y_true, y_prob, num_classes=num_classes)
fig_pr = plot_roc_curve_by_class(fpr, tpr, roc_auc, labels=labels)
all_plots['pr'] = wandb.Image(fig_pr)
plt.close(fig_pr)
return all_plots
"""
Highlights
==========
"""
def save_or_show_sample(plot_name):
"""
Add save/show
"""
def real_decorator(plot_func):
@wraps(plot_func)
def wrapper(*args, **kwargs):
save_to = kwargs.get("save_to", None)
model_name=kwargs.get("model_name", None)
index=kwargs.get("i", None)
fig = plot_func(*args, **kwargs)
if save_to is not None:
save_to = Path(save_to)
fig.savefig(save_to / f'{model_name}_{plot_name}_{index}.png')
show = kwargs.get("show", False)
if show:
plt.show()
plt.close()
return None
else:
return fig
return wrapper
return real_decorator
@save_or_show_sample("Prediction")
def plot_prediction(Xi, *args, **kwargs):
Xi = Xi.squeeze()
ndim = Xi.ndim
if ndim == 1:
# Signal 1D
fig = plt.figure(figsize=(5, 4))
plt.plot(Xi)
#plt.ylim([-1.0, 1.0])
plt.xlabel('$\Delta t$', fontsize=16)
plt.ylabel('Amplitude', fontsize=16)
plt.tight_layout()
plt.grid(True)
return fig
else:
raise NotImplementedError("Xi ndim > 1")
def get_highlights(X, y_true, y_prob, y_absolute_index, labels, top_k=10):
# Sort by performance
cross_entropy = -1*np.log(y_prob[np.arange(len(y_true)), y_true] + np.finfo(np.float32).eps)
sorted_index = np.argsort(cross_entropy)
# Select highlights
if y_absolute_index is None:
y_absolute_index = np.arange(len(y_true))
# Table data
## id
id_top_best, id_top_worst = list(y_absolute_index[sorted_index[:top_k]]), list(y_absolute_index[sorted_index[-top_k:]])
## target
y_true_top_best, y_true_top_worst = y_true[sorted_index[:top_k]], y_true[sorted_index[-top_k:]]
target_top_best, target_top_worst = [labels[yi] for yi in y_true_top_best], [labels[yi] for yi in y_true_top_worst]
## prediction
y_pred_top_best, y_pred_top_worst = y_prob.argmax(axis=-1)[sorted_index[:top_k]], y_prob.argmax(axis=-1)[sorted_index[-top_k:]]
prediction_top_best, prediction_top_worst = [labels[yi] for yi in y_pred_top_best], [labels[yi] for yi in y_pred_top_worst]
## X
X_top_best, X_top_worst = X[sorted_index[:top_k]], X[sorted_index[-top_k:]]
X_plot_top_best = []
for Xi in X_top_best:
fig = plot_prediction(Xi=Xi)
X_plot_top_best.append(wandb.Image(fig))
plt.close(fig)
X_plot_top_worst = []
for Xi in X_top_worst:
fig = plot_prediction(Xi=Xi)
X_plot_top_worst.append(wandb.Image(fig))
plt.close(fig)
## Cross-Entropy
cross_entropy_top_best, cross_entropy_top_worst = list(cross_entropy[sorted_index[:top_k]]), list(cross_entropy[sorted_index[-top_k:]])
# Buid Tables
columns=["id", "X", "target", "prediction", "cross_entropy"]
transpose_data_best = [
id_top_best, X_plot_top_best, target_top_best, prediction_top_best, cross_entropy_top_best
]
data_best = [list(row) for row in zip(*transpose_data_best)] #transpose
transpose_data_worst = [
id_top_worst, X_plot_top_worst, target_top_worst, prediction_top_worst, cross_entropy_top_worst
]
data_worst = [list(row) for row in zip(*transpose_data_worst)] #transpose
return {
"top_best": wandb.Table(columns=columns, data=data_best),
"top_worst": wandb.Table(columns=columns, data=data_worst)
}