-
Notifications
You must be signed in to change notification settings - Fork 13
/
ARD_NMF.py
executable file
·308 lines (273 loc) · 12.9 KB
/
ARD_NMF.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
import pandas as pd
import numpy as np
import sys
from sys import stdout
import argparse
import time
from scipy.special import gamma
import os
import pickle
import math
import torch
from typing import Union
import multiprocessing.connection as mpc
from .NMF_functions import *
class ARD_NMF:
"""
NMF results class implements both half normal and exponential prior ARD NMF
implementation based on https://arxiv.org/pdf/1111.6085.pdf
"""
def __init__(self,dataset,objective,dtype = torch.float32, verbose=True):
self.eps_ = torch.tensor(1.e-30,dtype=dtype,requires_grad=False)
self.dataset = dataset
zero_idx = np.sum(self.dataset, axis=1) > 0
self.V0 = self.dataset.values[zero_idx, :]
self.V = self.V0 - np.min(self.V0) + 1.e-30
self.V_max = np.max(self.V)
self.M = self.V.shape[0]
self.N = self.V.shape[1]
self.objective = objective
self.channel_names = self.dataset.index[zero_idx]
self.sample_names = self.dataset.columns
self.dtype = dtype
self.verbose = verbose
if self.verbose: print('NMF class initalized.')
def initalize_data(self,a,phi,b,prior_W,prior_H,Beta,K0,dtype = torch.float32):
"""
Initializes dataset.
Args:
* a
* phi: dispersion parameter - multiplied by variance if objective
function is Gaussian (see Tan & Fevotte 2013)
* b
* prior_W
* prior_H
* Beta
* K0: set to number of input features if not provided
"""
print('NMF class initialized.')
def initalize_data(self,a,phi,b,prior_W,prior_H,Beta,K0,use_val_set,dtype = torch.float32):
self.V = np.array(self.V) #when gets called in a loop as in run_parameter_sweep this can get updated to a torch tensor in a previous iteration which breaks some numpy functions
if K0 == None:
self.K0 = self.M
self.number_of_active_components = self.M
else:
self.K0 = K0
self.number_of_active_components = self.K0
if self.objective.lower() == 'poisson':
self.phi = torch.tensor(phi,dtype=dtype,requires_grad=False)
else:
self.phi = torch.tensor(np.var(self.V)* phi,dtype=dtype,requires_grad=False)
if use_val_set:
torch.manual_seed(0) #get the same mask each time
self.mask = (torch.rand(self.V.shape) > 0.2).type(self.dtype) #create mask, randomly mask ~20% of data in shape V. Only used when passed
else:
self.mask = torch.ones(self.V.shape, dtype=self.dtype)
self.a = a
self.prior_W = prior_W
self.prior_H = prior_H
self.C = []
self.b = b
W0 = np.multiply(np.random.uniform(size=[self.M, self.K0])+self.eps_.numpy(), np.sqrt(self.V_max))
H0 = np.multiply(np.random.uniform(size=[self.K0, self.N])+self.eps_.numpy(), np.sqrt(self.V_max))
L0 = np.sum(W0,axis=0) + np.sum(H0,axis=1)
self.W = torch.tensor(W0, dtype=self.dtype, requires_grad=False)
self.H = torch.tensor(H0, dtype=self.dtype, requires_grad=False)
self.Lambda = torch.tensor(L0, dtype=torch.float32, requires_grad=False)
# calculate default b as described in Tan and Fevotte (2012)
if self.b == None or self.b == 'None':
# L1 ARD
if self.prior_H == 'L1' and self.prior_W == 'L1':
self.bcpu = np.sqrt(np.true_divide( (self.a - 1)*(self.a - 2) * np.mean(self.V),self.K0 ))
self.b = torch.tensor(
np.sqrt(np.true_divide( (self.a - 1)*(self.a - 2) * np.mean(self.V),self.K0 ))
,dtype=self.dtype,requires_grad=False)
self.C = torch.tensor(self.N + self.M + self.a + 1, dtype=self.dtype, requires_grad=False)
# L2 ARD
elif self.prior_H == 'L2' and self.prior_W == 'L2':
self.bcpu = np.true_divide(np.pi * (self.a - 1) * np.mean(self.V),2*self.K0)
self.b = torch.tensor(
np.true_divide(np.pi * (self.a - 1) * np.mean(self.V),2*self.K0),
dtype=self.dtype,requires_grad=False)
self.C = torch.tensor( (self.N + self.M)*0.5 + self.a + 1, dtype=self.dtype,requires_grad=False)
# L1 - L2 ARD
elif self.prior_H == 'L1' and self.prior_W == 'L2':
self.bcpu = np.true_divide(np.mean(self.V)*np.sqrt(2)*gamma(self.a-3/2),self.K0*np.sqrt(np.pi)*gamma(self.a))
self.b = torch.tensor(
np.true_divide(np.mean(self.V)*np.sqrt(2)*gamma(self.a-3/2),self.K0*np.sqrt(np.pi)*gamma(self.a))
,dtype=self.dtype,requires_grad=False)
self.C = torch.tensor(self.N + self.M/2 + self.a + 1, dtype=self.dtype)
elif self.prior_H == 'L2' and self.prior_W == 'L1':
self.bcpu = np.true_divide(np.mean(self.V)*np.sqrt(2)*gamma(self.a-3/2),self.K0*np.sqrt(np.pi)*gamma(self.a))
self.b = torch.tensor(
np.true_divide(np.mean(self.V)*np.sqrt(2)*gamma(self.a-3/2),self.K0*np.sqrt(np.pi)*gamma(self.a)),
dtype=self.dtype,requires_grad=False)
self.C = torch.tensor(self.N/2 + self.M + self.a + 1, dtype=self.dtype)
else:
self.bcpu = self.b
self.b = torch.tensor(self.b, dtype=self.dtype,requires_grad=False)
if self.prior_H == 'L1' and self.prior_W == 'L1':
self.C = torch.tensor(self.N + self.M + self.a + 1, dtype=self.dtype,requires_grad=False)
# L2 ARD
elif self.prior_H == 'L2' and self.prior_W == 'L2':
self.C = torch.tensor( (self.N + self.M)*0.5 + self.a + 1, dtype=self.dtype,requires_grad=False)
# L1 - L2 ARD
elif self.prior_H == 'L1' and self.prior_W == 'L2':
self.C = torch.tensor(self.N + self.M/2 + self.a + 1, dtype=self.dtype,requires_grad=False)
elif self.prior_H == 'L2' and self.prior_W == 'L1':
self.C = torch.tensor(self.N/2 + self.M + self.a + 1, dtype=self.dtype,requires_grad=False)
self.V = torch.tensor(self.V,dtype=self.dtype,requires_grad=False)
if self.verbose: print('NMF data and parameters set.')
def get_number_of_active_components(self):
self.number_of_active_components = torch.sum(torch.sum(self.W,0)> 0.0, dtype=self.dtype)
def print_report(iter,report,verbose,tag):
"""
Prints report.
"""
if verbose:
print("nit={:>5} K={:>5} | obj={:.2f}\tb_div={:.2f}\tlam={:.2f}\tdel={:.8f}\tsumW={:.2f}\tsumH={:.2f}".format(
iter,
report[iter]['K'],
report[iter]['obj'],
report[iter]['b_div'],
report[iter]['lam'],
report[iter]['del'],
report[iter]['W_sum'],
report[iter]['H_sum']
)
)
else:
stdout.write("\r{}nit={:>5} K={} \tdel={:.8f}".format(
tag,
iter,
report[iter]['K'],
report[iter]['del']
)
)
def run_method_engine(
results: ARD_NMF,
a: float,
phi: float,
b: float,
Beta: int,
W_prior: str,
H_prior: str,
K0: int,
tolerance: float,
max_iter: int,
use_val_set: bool,
report_freq: int = 10,
active_thresh: float = 1e-5,
send_end: Union[mpc.Connection, None] = None,
cuda_int: Union[int, None] = 0,
verbose: bool = True,
tag: str = ""
) -> (pd.DataFrame, pd.DataFrame, np.ndarray, pd.DataFrame, np.ndarray):
"""
Run ARD-NMF Engine.
------------------------------------------------------------------------
Args:
* results: initialized ARD_NMF class
* a: shape parameter
* phi: dispersion parameter
* b: shape parameter
* Beta: defined by objective function
* W_prior: prior on W matrix ("L1" or "L2")
* H_prior: prior on H matrix ("L1" or "L2")
* K0: starting number of latent components
* tolerance: end-point of optimization
* max_iter: maximum number of iterations for algorithm
* use_val_set: use validation set for ARD-NMF
If False (default), set masks to all ones.
Otherwise, use 0/1 mask to hold out 0's as validation set during training and will report objective function value for that set.
* report_freq: how often to print updates
* active_thresh: threshold for a latent component's impact on
signature if the latent factor is less than this, it does not contribute
* send_end: mpc.Connection resulting from multiprocessing.Pipe,
for use in parameter sweep implementation
* cuda_int: GPU to use. Defaults to 0. If "None" or if no GPU available,
will perform decomposition using CPU.
* verbose: verbose logging
Returns:
* H: (samples x K)
* W: (K x features)
* markers
* signatures
"""
# initalize the NMF run
results.initalize_data(a, phi, b, W_prior, H_prior, Beta, K0, use_val_set)
# specify GPU
cuda_string = 'cuda:'+str(cuda_int)
# copy data to GPU
if torch.cuda.device_count() > 0 and cuda_int is not None:
if verbose: print(" * Using GPU: {}".format(cuda_string))
W,H,V,Lambda,C,b0,eps_,phi,mask = results.W.cuda(cuda_string),results.H.cuda(cuda_string),results.V.cuda(cuda_string),results.Lambda.cuda(cuda_string),results.C.cuda(cuda_string),results.b.cuda(cuda_string),results.eps_.cuda(cuda_string),results.phi.cuda(cuda_string),results.mask.cuda(cuda_string)
else:
W,H,V,Lambda,C,b0,eps_,phi,mask = results.W,results.H,results.V,results.Lambda,results.C,results.b,results.eps_,results.phi,results.mask
if verbose: print(" * Using CPU")
# tracking variables
deltrack = 1000
times = list()
report = dict()
iter = 0
lam_previous = Lambda
if verbose: print('%%%%%%%%%%%%%%%')
if verbose: print('a =',results.a)
if verbose: print('b =',results.bcpu)
if verbose: print('%%%%%%%%%%%%%%%')
# set method
method = NMF_algorithim(Beta, H_prior, W_prior)
start_time = time.time()
while deltrack >= tolerance and iter < max_iter:
# compute updates
H,W,Lambda = method.forward(W,H,V,Lambda,C,b0,eps_,phi,mask)
# compute objective and cost (excluding validation set, when mask is passed)
l_ = beta_div(Beta,V,W,H,eps_,mask)
cost_ = calculate_objective_function(Beta,V,W,H,Lambda,C,eps_,phi,results.K0,mask)
# update tracking
deltrack = torch.max(torch.div(torch.abs(Lambda-lam_previous), lam_previous+1e-30))
lam_previous = Lambda
# ---------------------------- Reporting ---------------------------- #
if iter % report_freq == 0:
report[iter] = {
'K': torch.sum((torch.sum(H,1) * torch.sum(W,0))>active_thresh).cpu().numpy(),
'obj': cost_.cpu().numpy(),
'b_div': l_.cpu().numpy(),
'lam': torch.sum(Lambda).cpu().numpy(),
'del': deltrack.cpu().numpy(),
'W_sum': torch.sum(W).cpu().numpy(),
'H_sum': torch.sum(H).cpu().numpy()
}
print_report(iter,report,verbose,tag)
# ------------------------------------------------------------------- #
iter+=1
# --------------------------- Final Report --------------------------- #
report[iter] = {
'K': torch.sum((torch.sum(H,1) * torch.sum(W,0))>active_thresh).cpu().numpy(),
'obj': cost_.cpu().numpy(),
'b_div': l_.cpu().numpy(),
'lam': torch.sum(Lambda).cpu().numpy(),
'del': deltrack.cpu().numpy(),
'W_sum': torch.sum(W).cpu().numpy(),
'H_sum': torch.sum(H).cpu().numpy()
}
end_time = time.time()
#compute validation set performance
if use_val_set:
heldout_mask = 1-mask #now select heldout values (inverse of mask)
report[iter]['b_div_val'] = beta_div(Beta,V,W,H,eps_,heldout_mask)
report[iter]['obj_val'] = calculate_objective_function(Beta,V,W,H,Lambda,C,eps_,phi,results.K0,heldout_mask)
#print("validation set objective=%s\tbeta_div=%s" % (cost_.cpu().numpy(),l_.cpu().numpy()))
else:
report[iter]['b_div_val'] = None
report[iter]['obj_val'] = None
print_report(iter,report,verbose,tag)
if not verbose:
stdout.write("\n")
# ------------------------------------------------------------------- #
if send_end != None:
send_end.send([W.cpu().numpy(), H.cpu().numpy(), mask.cpu().numpy(), cost_.cpu().numpy(), l_.cpu().numpy(), report[iter]['b_div_val'].cpu().numpy(), report[iter]['obj_val'].cpu().numpy(), end_time-start_time,])
else:
final_report = pd.DataFrame.from_dict(report).T
final_report.index.name = 'iter'
return W.cpu().numpy(), H.cpu().numpy(), cost_.cpu().numpy(), final_report, Lambda.cpu().numpy(), mask.cpu().numpy()