-
Notifications
You must be signed in to change notification settings - Fork 7
/
GeneNet.py
189 lines (148 loc) · 5.99 KB
/
GeneNet.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
import numpy as np
import matplotlib.pyplot as plt
import theano.tensor as T
import theano
from theano.tensor.shared_randomstreams import RandomStreams
from Model import dt
# theano.config.optimizer = "None"
###################################################################
# Optimizer
def AdamOptimizer(cost, params, lr=0.1, b1=0.02, b2=0.001, e=1e-8):
updates = []
grads = T.grad(cost, params)
i = theano.shared(np.float32(1))
i_t = i + 1.
fix1 = 1. - (1. - b1)**i_t
fix2 = 1. - (1. - b2)**i_t
lr_t = lr * (T.sqrt(fix2) / fix1)
for p, g in zip(params, grads):
m = theano.shared(np.zeros(p.get_value().shape, dtype=theano.config.floatX))
v = theano.shared(np.zeros(p.get_value().shape, dtype=theano.config.floatX))
m_t = (b1 * g) + ((1. - b1) * m)
v_t = (b2 * T.sqr(g)) + ((1. - b2) * v)
g_t = m_t / (T.sqrt(v_t) + e)
p_t = p - (lr_t * g_t)
updates.append((m, m_t))
updates.append((v, v_t))
updates.append((p, p_t))
updates.append((i, i_t))
return updates
####################################################################
# Define dynamics of the network using theano.scan
def evolveTime(initial, input_,network):
result, updates = theano.scan(
fn = lambda i, o: o + dt*network.networkFunction(o,i),
sequences = input_,
outputs_info = initial)
return result[-1][-1][:]
#returns final state
def fullEvolveTime(initial, input_,network):
result, updates = theano.scan(
fn = lambda i, o: o + dt*network.networkFunction(o,i),
sequences = input_,
outputs_info = initial)
return result
#returns entire time trace
####################################################################
# Define error function(s)
def costFunction(initial_, input_, output_,network):
# unregularized cost
y = network.normalize(evolveTime(initial_, input_,network))
y_ = output_
return T.sum(((y-y_)**2))
def costFunctionL1(initial_, input_, output_,network,L1):
# regularized cost
y = network.normalize(evolveTime(initial_, input_,network))
y_ = output_
L1norm = L1 * network.regularize()
return T.sum(((y-y_)**2)) + L1norm
####################################################################
# Execute training of network
def trainNetwork(desiredFunction, network, iterations, batchSize,mode = 'default',L1=0.1, pruneLimit = 1.0, plotCost = False):
#build computational graph
initial = T.matrix(name='initial',dtype=theano.config.floatX)
networkInput = T.tensor3(name='networkInput',dtype=theano.config.floatX)
desiredOutput = T.vector(name='desiredOutput',dtype=theano.config.floatX)
L1mask = T.matrix(name='L1mask',dtype=theano.config.floatX)
applyL1mask = theano.function(inputs = [L1mask],
updates = [(network.W, network.W*L1mask)])
simulationOutput = evolveTime(initial, networkInput,network)
simulate = theano.function(
inputs = [initial, networkInput],
outputs = simulationOutput)
fullSimulationOutput = fullEvolveTime(initial, networkInput,network)
fullSimulate = theano.function(
inputs = [initial, networkInput],
outputs = fullSimulationOutput)
cost = costFunction(initial, networkInput, desiredOutput,network)
train = theano.function(
inputs = [initial, networkInput, desiredOutput],
outputs = cost,
updates = AdamOptimizer(cost, network.parameters))
costL1 = costFunctionL1(initial, networkInput, desiredOutput,network,L1)
trainL1 = theano.function(
inputs = [initial, networkInput, desiredOutput],
outputs = costL1,
updates = AdamOptimizer(costL1, network.parameters))
# train network
costValue = []
iteration = 0
if mode is 'default':
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = train(initial, networkInput, desiredOutput)
costValue = np.append(costValue, current_cost.item())
iteration += 1
iteration = 0
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = trainL1(initial, networkInput, desiredOutput)
costValue = np.append(costValue, current_cost.item())
iteration += 1
mask = np.abs(network.W.eval()) > pruneLimit
iteration = 0
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = train(initial, networkInput, desiredOutput)
applyL1mask(mask)
costValue = np.append(costValue, current_cost.item())
iteration += 1
if mode is 'vanilla':
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = train(initial, networkInput, desiredOutput)
costValue = np.append(costValue, current_cost.item())
iteration += 1
if mode is 'regularize':
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = trainL1(initial, networkInput, desiredOutput)
costValue = np.append(costValue, current_cost.item())
iteration += 1
if mode is 'prune':
mask = np.abs(network.W.eval()) > pruneLimit
while (iteration < iterations):
[initial,networkInput,desiredOutput] = desiredFunction(batchSize)
current_cost = train(initial, networkInput, desiredOutput)
applyL1mask(mask)
costValue = np.append(costValue, current_cost.item())
iteration += 1
if (plotCost):
plt.plot(np.linspace(0,1,costValue.size),costValue)
plt.show()
return network.parameters
def simulateNetwork(inputData, network, batchSize):
initial = T.matrix(name='initial',dtype=theano.config.floatX)
networkInput = T.tensor3(name='networkInput',dtype=theano.config.floatX)
desiredOutput = T.vector(name='desiredOutput',dtype=theano.config.floatX)
simulationOutput = evolveTime(initial, networkInput,network)
simulate = theano.function(
inputs = [initial, networkInput],
outputs = simulationOutput)
fullSimulationOutput = fullEvolveTime(initial, networkInput,network)
fullSimulate = theano.function(
inputs = [initial, networkInput],
outputs = fullSimulationOutput)
[initial,networkInput,desiredOutput] = inputData(batchSize)
outputSimulation = fullSimulate(initial, networkInput)
return outputSimulation