-
Notifications
You must be signed in to change notification settings - Fork 0
/
SFA_Node.py
168 lines (134 loc) · 6.05 KB
/
SFA_Node.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
from mdp.nodes import TimeFramesNode, PolynomialExpansionNode
import numpy as np
from IncSFA.incsfa import IncSFANode
from IncSFA.trainer import TrainerNode
class WindowDelayNode:
def __init__(self, past_samples=1):
self.past_samples = past_samples
self.delaynode = TimeFramesNode(self.past_samples)
def execute(self,x):
if x.ndim == 3:
return np.stack([self.delaynode.execute(win) for win in x], axis=0, out=None)
elif x.ndim == 2:
return self.delaynode.execute(x)
elif x.ndim == 1:
return self.delaynode.execute(np.expand_dims(x, axis=1))
else:
raise ValueError(f"Invalid input {x.shape=} (max 3 dimensions).")
return None
def __str__(self):
return f"WindowDelayNode: TimeFramesNode(past_samples={self.past_samples})"
def get_past_samples(x,past_samples):
delaynode = WindowDelayNode(past_samples)
return delaynode.execute(x)
class SFA_Node:
def __init__(self, iterval=1, degree=1, past_samples=1, deMean=True, whitening_dim=None, output_dim=None, verbose_func=print, mode='Incremental'):
assert mode in ['Incremental', 'BlockIncremental', 'Batch']
self.__dict__.update(locals())
del self.__dict__['self']
self.node = None
self.trainer = None
self.delaynode = WindowDelayNode(self.past_samples)
self.expnode = PolynomialExpansionNode(self.degree)
if self.verbose_func is not None:
self.verbose_func(f"Initialised SFA Node: {self}")
def train(self,data):
input_data = self.delaynode.execute(data)
if self.trainer is None and self.verbose_func is not None:
self.verbose_func(f"Train Poly degree {self.degree}: {data.shape=} {input_data.shape=}")
if data.ndim == 3:
for win in input_data:
self.train_node(win)
elif data.ndim == 2:
self.train_node(input_data)
else:
raise ValueError(f"Invalid input {data.shape=} {input_data.shape=} (must be 2 or 3 dimensions).")
def train_node(self,input_data):
input_data = self.expnode(input_data)
if self.trainer is None and self.verbose_func is not None:
self.verbose_func(f"Polynomial Expansion: {input_data.shape=}")
if self.node is None:
if self.whitening_dim is None: self.whitening_dim = input_data.shape[1]
if self.output_dim is None: self.output_dim = input_data.shape[1]
self.node = IncSFANode(input_dim=input_data.shape[1],
whitening_output_dim=self.whitening_dim,
output_dim=self.output_dim,
deMean=self.deMean,
eps=0.05)
if self.trainer is None:
self.trainer = TrainerNode(self.node, mode=self.mode, progressbar=False)
self.trainer.train(input_data, iterval=self.iterval)
def apply(self,data):
if self.node is None:
from warnings import warn
warn(f"SFA node is untrained! Will be trained on {data.shape=}")
self.train(data)
input_data = self.delaynode.execute(data)
#print("delayed data:")
#print(input_data.T)
if data.ndim == 3:
return np.stack([self.execute(win) for win in input_data], axis=0, out=None)
elif data.ndim == 2:
return self.execute(input_data)
else:
raise ValueError(f"Invalid input {data.shape=} {input_data.shape=} (must be 2 or 3 dimensions).")
def execute(self,input_data):
input_data = self.expnode(input_data)
#print("expanded:")
#print(input_data.T)
return self.node.execute(input_data)
def fit_transform(self,data):
del_data = self.delaynode.execute(data)
if data.ndim == 3:
win_ = []
for win in del_data:
exp_data = self.expnode(win)
self.fit_node(exp_data)
win_.append(self.node.execute(exp_data))
return np.stack(win_, axis=0, out=None)
elif data.ndim == 2:
exp_data = self.expnode(del_data)
self.fit_node(exp_data)
return self.node.execute(exp_data)
else:
raise ValueError(f"Invalid input {data.shape=} {del_data.shape=} (must be 2 or 3 dimensions).")
def fit(self,data):
pass
def transform(self,data):
pass
def fit_node(self,input_data):
if self.node is None:
if self.whitening_dim is None: self.whitening_dim = input_data.shape[1]
if self.output_dim is None: self.output_dim = input_data.shape[1]
self.node = IncSFANode(input_dim=input_data.shape[1],
whitening_output_dim=self.whitening_dim,
output_dim=self.output_dim,
deMean=self.deMean,
eps=0.05)
if self.trainer is None:
self.trainer = TrainerNode(self.node, mode=self.mode, progressbar=False)
self.trainer.train(input_data, iterval=self.iterval)
def __str__(self):
return str(self.__dict__)
if __name__ == '__main__':
import numpy.matlib
x1 = np.matlib.repmat([1, -1],1,50)
x2 = np.matlib.repmat([1, 0, -1, 0],1,25)
x3 = np.matlib.repmat([1, 2],1,50)
# x = np.expand_dims(np.concatenate((x1,
# x2,
# x3),
# axis = 0), axis=2)
# print(x.shape)
# x_ = get_past_samples(x,past_samples=2)
# print(x_)
# print(x_.shape)
sfa = SFA_Node(iterval=1, degree=2, past_samples=5, whitening_dim=1, output_dim=1, verbose_func=print, mode='Incremental')
x = np.concatenate((x1,x2,x3),axis = 0)
x = np.transpose(x)
print(x.shape)
#x = np.expand_dims(x, axis=0)
sfa.train(x)
out = sfa.apply(x)
print(out)
print(out.shape)