-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathoutput_layer.py
152 lines (129 loc) · 7.31 KB
/
output_layer.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
from copy import copy
import numpy as np
import pandas as pd
from sklearn import linear_model
from sklearn.linear_model import LassoCV
import booleanNetwork as bn
from reservoir import Reservoir, addReservoirParameters
from function_arguments import generateFunctionDataFromFunction
class OutputLayer(object):
def __init__(self, reservoir, numberOfOutputNodes, functions,
inputsToFunctions, delay, dataStreamLength, nonRecursiveArgs=[[]]):
# print('Beginning __init__')
n = numberOfOutputNodes
if len(functions) != n or len(inputsToFunctions) != n:
raise ValueError('Function parameter(s) do not match number of output nodes.')
self.reservoir = reservoir
self.inputData = np.array(self.reservoir.inputData, dtype=np.int8)
if self.inputData.ndim == 1:
self.inputData.shape = (len(self.inputData), 1)
self.numberOfOutputNodes = numberOfOutputNodes
self.functions = functions
self.inputsToFunctions = inputsToFunctions
self.delay = delay
self.dataStreamLength = dataStreamLength
if nonRecursiveArgs == [[]]:
self.nonRecursiveArgs = [[] for _ in range(numberOfOutputNodes)]
else:
self.nonRecursiveArgs = nonRecursiveArgs
self.numberOfInputsToFunctions = max(len(x) for x in inputsToFunctions)
self.windows = np.zeros(self.numberOfOutputNodes, dtype=int)
for i in range(self.numberOfOutputNodes):
self.windows[i] = 1 + max(t for (n, t) in inputsToFunctions[i])
if len(self.nonRecursiveArgs[i]) > 0: # fix this
nr_window = 1 + max(t for (n, t) in nonRecursiveArgs[i])
self.windows[i] = max(self.windows[i], nr_window)
self.window = max(self.windows)
# TODO: check the next two lines
self.shift = self.window - 1 + self.delay # I don't know if delay should be here
self.totalStreamLength = self.dataStreamLength + self.shift
self.inputDataIndex = 0
self.successRates = []
def generateData(self, numberOfDataStreams):
if numberOfDataStreams * self.totalStreamLength \
* self.reservoir.numberOfInputs > \
len(self.inputData) - self.inputDataIndex:
raise ValueError('Not enough input data to complete training/testing.')
# Set the values of all of the training inputs from the random input streams
trainingInput = np.full((numberOfDataStreams, self.totalStreamLength,
self.reservoir.numberOfInputs), -1, dtype=np.int8)
for i in range(numberOfDataStreams):
for j in range(self.totalStreamLength):
# this is a 1-D array with length self.reservoir.numberOfInputs
trainingInput[i,j] = self.inputData[i*self.totalStreamLength+j + self.inputDataIndex]
self.inputDataIndex += numberOfDataStreams * self.totalStreamLength * self.reservoir.numberOfInputs
if -1 in trainingInput:
raise ValueError('The the full training input was not initialized.')
# print('Input set trained.')
# Generate function data
trainingOutput = np.full((self.numberOfOutputNodes, numberOfDataStreams,
self.dataStreamLength), -1, dtype=np.int8)
for i in range(self.numberOfOutputNodes):
for j in range(numberOfDataStreams):
functionData = generateFunctionDataFromFunction(
trainingInput[j], self.windows[i],
self.inputsToFunctions[i], self.functions[i],
self.dataStreamLength,
nonRecursiveArgs=self.nonRecursiveArgs[i])
for k in range(len(functionData)):
trainingOutput[i,j,k] = functionData[k]
if -1 in trainingOutput:
raise ValueError('The the full training output was not initialized.')
# print('Output set trained.')
# generate Boolean network data
trainingNetwork = np.full((numberOfDataStreams, self.totalStreamLength,
self.reservoir.numberOfNetworkNodes),
-1, dtype=np.int8)
for i in range(numberOfDataStreams):
r_temp = copy(self.reservoir) # make sure this works and that I don't need to use deepcopy
initialState = bn.getRandomInitialNodeValues(self.reservoir.numberOfNetworkNodes)
r_temp.setInitialNodeValues(initialState)
r_temp.inputData = trainingInput[i]
r_temp.update(self.totalStreamLength - 1) # - 1 since we include the initial state
trainingNetwork[i] = r_temp.networkHistory[:,:-self.reservoir.numberOfInputs]
# print('Network data gathered.')
# generate X matrix
X_data = np.full((numberOfDataStreams * self.dataStreamLength,
self.reservoir.numberOfNetworkNodes),
-1, dtype=np.int8) # no type
for i in range(len(trainingNetwork)):
X_data[i * self.dataStreamLength:(i+1) * self.dataStreamLength] = trainingNetwork[i,self.shift:]
# Generate list of y-vectors
y_data = np.full((self.numberOfOutputNodes, numberOfDataStreams *
self.dataStreamLength), -1, dtype=np.int8) # no type
for i in range(self.numberOfOutputNodes):
for j in range(numberOfDataStreams):
y_data[i,j*self.dataStreamLength:(j+1) * self.dataStreamLength] = trainingOutput[i,j]
return (X_data, y_data)
def train(self, trainingSize):
# print('Training sequence initiated.')
X_train, y_train = self.generateData(trainingSize)
self.X_train = X_train
self.models = []
for i in range(self.numberOfOutputNodes):
reg = LassoCV(cv=5, max_iter=10000)
self.models.append(reg)
self.models[i].fit(X_train, y_train[i])
def test(self, testSize):
# print('Testing sequence initiated.')
X_test, y_test = self.generateData(testSize)
shifted_signum = lambda x : 1 if x > 0.5 else 0
y_predicted_raw = np.full((self.numberOfOutputNodes, testSize * self.dataStreamLength), -1.0)
y_predicted = np.full((self.numberOfOutputNodes, testSize * self.dataStreamLength), -1, dtype=np.int8)
for i in range(self.numberOfOutputNodes):
y_predicted_raw[i] = self.models[i].predict(X_test)
y_predicted[i] = np.array(list(map(shifted_signum, y_predicted_raw[i])))
differenceVector = [abs(y_test[i,j] - y_predicted[i,j]) for j in range(testSize * self.dataStreamLength)]
successRate = 1 - (sum(differenceVector) / (testSize * self.dataStreamLength))
self.successRates.append(successRate)
#y_predicted_raw = np.matmul(X_test, self.models[0].coef_)
# print('number of output nodes: %d' % self.numberOfOutputNodes)
# print(f'model weights: {self.models[0].coef_}')
# print('X_test, y_test:, y_predicted_raw, y_predicted')
# for i in range(len(X_test)):
# print(f'{X_test[i]} {y_test[0,i]} {y_predicted_raw[0,i]} {y_predicted[0,i]}')
# print('\n\n\n\n\n')
return (y_test, y_predicted)
def getSuccessRates(self):
return self.successRates
# Make sure to fix filepaths