-
Notifications
You must be signed in to change notification settings - Fork 0
/
brute_force.py
57 lines (48 loc) · 2.14 KB
/
brute_force.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
#!/usr/bin/env python3
"""
DEPRECATED: try using the new "two_tone" instead, still uses simple neural net and FFBP,
however, now offers two distinct output nodes for error detection
"""
# Rough outline of brute-force binary classification
# Issues: classification attempts to get ouput as 0 or 1, may not be clear cut
# thresholding function might be required
# Essentially- extracting inputs in form of k-mer count from sequence,
# using single perceptron classification to attempt feature detection
# Next build- attempt two perceptrons for output detection- one for each species
import numpy as np
class network:
def __init__(self):
self.hidden = [ 0 for i in range((4**10)+1)]
self.output = 0
self.bias = 0
self.weights = [ 1/(4**10) for i in range((4**10)+1)]
self.error = 0
self.delta = 0
self.delta_weights = [0 for i in range((4**10)+1)]
def feed_forward(self, sequence):
for i in range(len(sequence)-9):
add_index = get_index(sequence[i:i+10])
self.hidden[add_index] += 1
for i in range((4**10)+1): output += self.hidden[i] * self.weights[i]
self.output = 1/(1+ np.exp(-output))
def back_propagation(self, species):
if species == "HIV": target = 1
else: target = 0
self.error = target - self.output
self.delta = self.error * (1 - self.output) * self.output
def update_weights(self, eta):
self.bias = slef.bias + self.delta * eta
self.delta_weights = [ self.delta * eta * self.hidden[i] for i in range((4**10)+1)]
self.weights = [ self.weights[i] + self.delta_weights[i] for i in range((4**10)+1)]
def report(self,sequence):
self.feed_forward(sequence)
return self.output
def get_index(sequence):
index = 0
for i in range(len(sequence)):
token = sequence[len(sequence)-1-i:len(sequence)-i]
if token == "A": index += (0 * (4**i))
elif token == "C": index += (1 * (4**i))
elif token == "G": index += (2 * (4**i))
else: index += (3 * (4**i))
return index