-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
394 lines (328 loc) · 13.3 KB
/
dataset.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
386
387
388
389
390
391
392
393
394
from torchvision.datasets import MNIST
from torch.utils.data import Dataset
import numpy as np
import torch
from sklearn import preprocessing
from scipy.io import loadmat
from sklearn.preprocessing import MinMaxScaler
from scipy.stats import zscore
import matplotlib.pyplot as plt
from sklearn import datasets
class ClusteringDataset(Dataset):
def __init__(self, data, labels=None, num_clusters=None):
super().__init__()
self.data = data
self.labels = labels
self._num_clusters = num_clusters
if num_clusters is None and labels is None:
raise ValueError("At least one of the values should be provided (labels/num_clusters)")
self.print_stats()
def __getitem__(self, index: int):
if self.labels is None:
return torch.tensor(self.data[index]).float()
return torch.tensor(self.data[index]).float(), torch.tensor(self.labels[index]).long()
def __len__(self) -> int:
return len(self.data)
@property
def num_clusters(self):
return self._num_clusters if self._num_clusters is not None else len(np.unique(self.labels))
def num_features(self):
return self.data.shape[-1]
def print_stats(self):
print('X.shape: ', self.data.shape)
print(f"X.min={self.data.min()}, X.max={self.data.max()}")
if self.labels is not None:
print('Y.shape: ', self.labels.shape)
for y_u in np.unique(self.labels):
print(f'{y_u}: {np.sum(self.labels == y_u)}')
print(f"Y.min={self.labels.min()}, Y.max={self.labels.max()}")
@classmethod
def setup(cls, cfg):
pass
class PBMC(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
data_dir = cfg.data_dir
with np.load(f"{data_dir}/pbmc_x.npz") as data:
X = data['arr_0']
with np.load(f"{data_dir}/pbmc_y.npz") as data:
Y = data['arr_0']
Y = Y - Y.min()
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class BIASE(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
name = 'biase'
data_dir = cfg.data_dir
dataset_x = f"{data_dir}/{name}/{name}_data.csv"
dataset_y = f"{data_dir}/{name}/{name}_celldata.csv"
with open(dataset_x) as r:
data = [l.strip() for l in r.readlines()]
cell_keys = data[0].split(',')[1:]
rows = [np.array([float(v) for v in row.split(',')[1:]]).reshape((1, -1)) for row in data[1:]]
X = BIASE.remove_zero_columns(
np.concatenate(rows, axis=0).transpose()) # np.concatenate(rows, axis=0).transpose()
with open(dataset_y) as r:
y_data = [l.strip().split(',') for l in r.readlines()[1:]]
cell2class = {row[0]: row[2] for row in y_data}
class2count = {}
for cell, clas in cell2class.items():
class2count.setdefault(clas, 0)
class2count[clas] += 1
print(class2count)
class2id = {c: i for i, c in enumerate(set(sorted(list(cell2class.values()))))}
Y = []
for cell_key in cell_keys:
Y.append(class2id[cell2class[cell_key]])
Y = np.array(Y).reshape(-1)
X = BIASE.transform(X)
X = np.log(1 + X)
X = X + .001 * np.random.normal(0, 1, (X.shape))
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class INTESTINE(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
scaler = getattr(preprocessing, cfg.scaler)()
name = 'intestine'
data_dir = cfg.data_dir
dataset_x = f"{data_dir}/{name}/{name}_data.csv"
dataset_y = f"{data_dir}/{name}/{name}_celldata.csv"
with open(dataset_x) as r:
data = [l.strip() for l in r.readlines()]
cell_keys = data[0].split(',')[1:]
rows = [np.array([float(v) for v in row.split(',')[1:]]).reshape((1, -1)) for row in data[1:]]
X = np.concatenate(rows, axis=0).T
with open(dataset_y) as r:
y_data = [l.strip().split(',') for l in r.readlines()[1:]]
cell2class = {row[0]: row[2] for row in y_data}
class2count = {}
for cell, clas in cell2class.items():
class2count.setdefault(clas, 0)
class2count[clas] += 1
print(class2count)
class2id = {c: i for i, c in enumerate(sorted(set(list(cell2class.values()))))}
Y = []
for cell_key in cell_keys:
Y.append(class2id[cell2class[cell_key]])
Y = np.array(Y).reshape(-1)
X = scaler.fit_transform(X)
return cls(X, Y)
class CNAE9(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
scaler = getattr(preprocessing, cfg.scaler)()
data = np.loadtxt(f"{cfg.data_dir}/cnae_9_numpy.txt")
X = data[:, :-1]
Y = data[:, -1]
Y = Y - Y.min()
X = scaler.fit_transform(X)
return cls(X, Y)
class MFEATZERNIKE(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
scaler = getattr(preprocessing, cfg.scaler)()
data = np.loadtxt(f"{cfg.data_dir}/mfeat_zernike_numpy.txt")
X = data[:, :-1]
Y = data[:, -1]
Y = Y - Y.min()
X = scaler.fit_transform(X)
return cls(X, Y)
class ALLAML(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
dataset = loadmat(f"{cfg.data_dir}/ALLAML.mat")
X = dataset.get('X')
Y = dataset.get('Y').reshape(-1)
Y = Y - Y.min()
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class PROSTATE(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
dataset = loadmat(f"{cfg.data_dir}/PROSTATE.mat")
X = dataset.get('X')
Y = dataset.get('Y').reshape(-1)
Y = Y - Y.min() # to start from zero
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class TOX171(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
dataset = loadmat(f"{cfg.data_dir}/TOX171.mat")
X = dataset.get('X')
Y = dataset.get('Y').reshape(-1)
Y = Y - Y.min() # to start from zero
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class SRBCT(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
dataset = loadmat(f"{cfg.data_dir}/SRBCT.mat")
X = dataset.get('X')
Y = dataset.get('Y').reshape(-1)
Y = Y - Y.min() # to start from zero
scaler = getattr(preprocessing, cfg.scaler)()
X = scaler.fit_transform(X)
return cls(X, Y)
class MNIST60K(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
scaler = getattr(preprocessing, cfg.scaler)()
X = MNIST(cfg.data_dir, train=True, download=True).data.reshape(-1, 784).cpu().numpy()
Y = MNIST(cfg.data_dir, train=True, download=True).targets.cpu().numpy()
X = scaler.fit_transform(X)
return cls(X, Y)
class MNIST10K(ClusteringDataset):
def __init__(self, data, targets):
super().__init__(data, targets)
@classmethod
def setup(cls, cfg):
scaler = getattr(preprocessing, cfg.scaler)()
X = MNIST(cfg.data_dir, train=True, download=True).data.reshape(-1, 784).cpu().numpy()
Y = MNIST(cfg.data_dir, train=True, download=True).targets.cpu().numpy()
X = scaler.fit_transform(X)
X = X[:10000]
Y = Y[:10000]
return cls(X, Y)
class NumpyTableDataset(ClusteringDataset):
def __init__(self, data, labels=None, num_clusters=None):
super().__init__(data, labels, num_clusters)
@classmethod
def setup(cls, filepath_samples: str, filepath_labels: str = None, num_clusters: int = None):
"""
:param filepath_samples: the path to the npz file, the format of the numpy array should be NxD
(number of samples x number of features)
:param filepath_labels: the path to the npz file, the format of the numpy array should be N
(number of samples)
:param num_clusters: the integer number of expected clusters
"""
with np.load(filepath_samples) as data:
X = data['arr_0']
if filepath_labels is not None:
with np.load(filepath_labels) as data:
Y = data['arr_0']
X = preprocessing.StandardScaler().fit_transform(X)
Y = Y - Y.min()
else:
Y = None
return cls(X, Y, num_clusters)
def remove_zero_columns(X):
non_zero_columns = []
for col in range(X.shape[1]):
if np.min(X[:, col]) == 0 and np.max(X[:, col]) == 0:
continue
else:
non_zero_columns.append(col)
X = X[:, non_zero_columns]
return X
class Synthetic(Dataset):
def __init__(self, X, Y):
super().__init__()
self.data = X
self.targets = Y
def __getitem__(self, index: int):
x = self.data[index]
return torch.tensor(x).float(), torch.tensor(self.targets[index]).long()
def __len__(self) -> int:
return len(self.data)
@classmethod
def setup(cls, num_samples=5000, num_features=3, num_clusters=3, num_noise_dims=10):
"""
Make num_clusters + 1 clusters in 3d and adds additional num_noise_dims noise features
:param num_samples: number of samples in the dataset
:param num_features: number of features in the dataset
:param num_clusters: number of clusters in the dataset
:param num_noise_dims: number of noise dimensions in addition to num_features
:return: generates a dataset
"""
x_2d, y_2d = datasets.make_blobs(num_samples, num_features-1, centers=num_clusters, cluster_std=.5,
random_state=0)
# split the points for cluster==2 into 2 clusters:
max_x = x_2d[:, 1].max()
min_x = x_2d[:, 1].min()
x_y_2 = x_2d[y_2d == 2][:, 1]
x_y_2 = MinMaxScaler((0, 1)).fit_transform(x_y_2.reshape(-1, 1)).reshape(-1)
x_y_2 = MinMaxScaler((min_x, max_x)).fit_transform(x_y_2.reshape(-1, 1)).reshape(-1)
x_2d[:, 1][y_2d == 2] = x_y_2
z = np.random.rand(num_samples)
y_2d[(y_2d == 2) & (z > 0.5)] = 3
x_2d[:, 0][y_2d == 0] = x_2d[:, 0][y_2d == 1]
bg = np.random.normal(loc=0, scale=0.01, size=(num_samples, num_noise_dims))
X = np.concatenate([x_2d, z.reshape(-1, 1), bg], axis=1)
X[:, 2][y_2d == 3] = X[:, 2][y_2d == 3] + 0.5 # separate in z axis
X[:, 2][y_2d == 0] = MinMaxScaler(
(X[:, 2][(y_2d == 3) | (y_2d == 2)].min(), X[:, 2][(y_2d == 3) | (y_2d == 2)].max())).fit_transform(
X[:, 2][y_2d == 0].reshape(-1, 1)).reshape(-1)
X[:, 2][y_2d == 1] = MinMaxScaler(
(X[:, 2][(y_2d == 3) | (y_2d == 2)].min(), X[:, 2][(y_2d == 3) | (y_2d == 2)].max())).fit_transform(
X[:, 2][y_2d == 1].reshape(-1, 1)).reshape(-1)
Y = y_2d
X4 = X[Y == 3]
max_len = len(X4)
X1 = X[Y == 0][:max_len, :]
X2 = X[Y == 1][:max_len, :]
X3 = X[Y == 2][:max_len, :]
Y1 = Y[Y == 0][:max_len]
Y2 = Y[Y == 1][:max_len]
Y3 = Y[Y == 2][:max_len]
Y4 = Y[Y == 3][:max_len]
X = np.concatenate([X1, X2, X3, X4], axis=0)
Y = np.concatenate([Y1, Y2, Y3, Y4], axis=0)
print("Class stats:")
for y_i in np.unique(Y):
print(f"{y_i}: {len(Y[Y == y_i])} samples")
X[:, :3] = zscore(X[:, :3])
plt.style.use('classic')
plt.rcParams['axes.spines.right'] = False
plt.rcParams['axes.spines.top'] = False
fig = plt.figure()
fig.set_facecolor('w')
plt.scatter(X[:, 0], X[:, 1], c=Y, s=100, alpha=0.8, cmap='viridis', edgecolor='k', linewidth=2)
plt.xlabel('$X_1$', fontsize=30)
plt.ylabel('$X_2$', fontsize=30)
plt.tight_layout()
plt.xticks([])
plt.yticks([])
plt.savefig("synth_X_1_X_2.png")
plt.clf()
fig = plt.figure()
fig.set_facecolor('w')
plt.scatter(X[:, 0], X[:, 2], c=Y, s=100, alpha=0.8, cmap='viridis', edgecolor='k', linewidth=2)
plt.xlabel('$X_1$', fontsize=30)
plt.ylabel('$X_3$', fontsize=30)
plt.tight_layout()
plt.xticks([])
plt.yticks([])
plt.savefig("synth_X_1_X_3.png")
return cls(X, Y)
def num_classes(self):
return len(np.unique(self.targets))
def num_features(self):
return self.data.shape[-1]