-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathFindHypers.py
171 lines (155 loc) · 7.09 KB
/
FindHypers.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
import numpy as n
import aipy as a
import pylab as pl
import xrfi
import torch
import torch.nn as nn
from torch.autograd import Variable
from glob import glob
import torchvision.datasets as dsets
import torchvision.transforms as transforms
from torch.utils.data import TensorDataset
from torch.utils.data import DataLoader
from sklearn.metrics import log_loss
def loadFullDay():
HERAlist = glob('/users/jkerriga/data/jkerriga/HERA/zen.2457458.*.xx.HH.uvcUA')
HERAdata = []
times = []
for k in HERAlist:
uvHERA = a.miriad.UV(k)
a.scripting.uv_selector(uvHERA, '9_10', 'xx')
#temp = []
for p,d,f in uvHERA.all(raw=True):
#temp.append(d)
HERAdata.append(d)
times.append(uvHERA['lst'])
#HERAdata.append(temp)
HERAdata = n.array(HERAdata)
times = n.array(times)
return HERAdata,times
def injectRandomRFI(data,mask,injections):
sh = n.shape(data)
print sh
outD = n.copy(data)
outM = n.copy(mask)
for i in range(injections):
if n.abs(n.random.rand())>0.5:
## RFI across time
fw = n.random.randint(1,3)+1
th = n.random.randint(1,900)+1
fs = n.random.randint(1,sh[1]-fw)
ts = n.random.randint(1,sh[0]-th)
outD[ts:ts+th,fs:fs+fw] = outD[ts:ts+th,fs:fs+fw]+0.01*n.random.randn()*(n.random.randn(th,fw)+ 1j*n.random.randint(-1,2)*n.random.randn(th,fw))
outM[ts:ts+th,fs:fs+fw] = 0.
else:
## RFI across freq
fw = n.random.randint(1,100)+1
th = n.random.randint(1,3)+1
fs = n.random.randint(1,sh[1]-fw)
ts = n.random.randint(1,sh[0]-th)
outD[ts:ts+th,fs:fs+fw] = outD[ts:ts+th,fs:fs+fw]+0.01*n.random.randn()*(n.random.randn(th,fw)+ 1j*n.random.randint(-1,2)*n.random.randn(th,fw))
outM[ts:ts+th,fs:fs+fw] = 0.
del(fw)
del(th)
return outD,outM
class CNN(nn.Module):
def __init__(self,dropRate,kernel):
self.dropRate = dropRate
self.kernel = kernel
super(CNN, self).__init__()
self.layer1 = nn.Sequential(
nn.BatchNorm2d(1),
nn.Conv2d(1, 16*1, kernel_size=self.kernel, padding=(self.kernel-1)/2),
nn.BatchNorm2d(16*1),
nn.MaxPool2d(kernel_size=(1,16)),
nn.ReLU())
self.layer2 = nn.Sequential(
nn.Conv2d(1*16, 2*16, kernel_size=self.kernel, padding=(self.kernel-1)/2),
nn.BatchNorm2d(2*16),
nn.MaxPool2d(kernel_size=(2,1)),
nn.ReLU())
#### Dropout Layer ####
self.layer3 = nn.Sequential(
nn.Conv2d(2*16, 2*16, kernel_size=self.kernel, padding=(self.kernel-1)/2),
nn.BatchNorm2d(2*16),
nn.Dropout2d(p=self.dropRate),
nn.ReLU())
self.layer4 = nn.Sequential(
nn.Linear(1024,1024),
nn.ReLU())
self.layer5 = nn.Sequential(
nn.Linear(1024,1024),
nn.ReLU())
self.layer6 = nn.Sequential(
nn.Linear(1024,1024),
nn.ReLU())
self.fc = nn.Linear(1024, 2*1024)
def forward(self, x):
out = self.layer1(x).view(1,16,-1,1024/16)
out = self.layer2(out).view(1,2*16,-1,1024/16)
out = self.layer3(out).view(-1,1024).float()
out = self.layer4(out).view(-1,1024).float()
out = self.layer5(out).view(-1,1024).float()
out = self.layer6(out).view(-1,1024).float()
out = self.fc(out)
return out
# Hyper Parameters
num_epochs = 1
batch_size = 1
grid = 5
lr = n.logspace(-5,-3,grid)
dr = n.linspace(0,0.9,grid)
ke = n.arange(1,11+grid,2)
LossGrid = n.zeros((grid,grid,grid))
###### Use a cost function to determine hyperparameters #####
for d in range(grid):
print d
for l in range(grid):
for k in range(grid):
DATA,times = loadFullDay()
DATA = DATA[0:500,:]
XRFImask = xrfi.xrfi_simple(DATA)
mask = n.loadtxt('trainMask_HQ.txt')[0:500,:]*n.logical_not(XRFImask[0:500,:]).astype(int)
#DATA,mask = injectRandomRFI(DATA,mask,int(injections))
MASK1 = n.array(mask).reshape(10,-1,1024,1)
DATA1 = DATA
DATA1 = n.abs(DATA)
DATA1 = DATA1/n.max(DATA1)
DATA1 = DATA1.reshape(10,-1,1024,1)
data1 = torch.from_numpy(DATA1)
mask1 = torch.from_numpy(MASK1)
train_dataset = TensorDataset(data1,mask1)
train_loader = DataLoader(train_dataset,batch_size=batch_size,shuffle=True)
test_loader = DataLoader(train_dataset,batch_size=batch_size,shuffle=False)
cnn = CNN(dr[d],ke[k])
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(cnn.parameters(), lr=lr[l])
# Train the Model
for epoch in range(num_epochs):
for i, (images, labels) in enumerate(train_loader):
images = Variable(images).float()
labels = Variable(labels[:,:,:,0]).long().view(-1)
# Forwar d + Backward + Optimize
optimizer.zero_grad()
images = images.view(1,1,-1,1024)
outputs = cnn(images)
outputs = outputs.view(-1,2)
loss = criterion(outputs.float(), labels.long())
loss.backward()
optimizer.step()
_, predicted = torch.max(outputs.data, 1)
print 'Evaluating...'
cnn.eval()
ll = 0.
for i, (images, labels) in enumerate(test_loader):
images = Variable(images).float()
images = images.view(1,1,-1,1024)
outputs = cnn(images)
outputs = outputs.view(-1,2)
_, singPred = torch.max(outputs.data, 1)
labels = labels.float().view(-1)
ll += log_loss(labels.numpy(),singPred.numpy())
LossGrid[d,l,k] = ll
bestFit = n.argwhere(LossGrid==n.min(LossGrid))
bF = bestFit[0]
print LossGrid[bF[0],bF[1],bF[2]],':','Best Drop Rate: ',dr[bF[0]],'Best Learning Rate: ',lr[bF[1]],'Best Kernel Size: ',ke[bF[2]]