Source Code for Module ML.Neural.ActFuncs

 1  # 
 2  #  Copyright (C) 2000  greg Landrum 
 3  # 
 4  """ Activation functions for neural network nodes 
 5   
 6  Activation functions should implement the following API: 
 7   
 8   - _Eval(input)_: returns the value of the function at a given point 
 9   
10   - _Deriv(input)_: returns the derivative of the function at a given point 
11   
12  The current Backprop implementation also requires: 
13   
14   - _DerivFromVal(val)_: returns the derivative of the function when its 
15                          value is val 
16   
17  In all cases _input_ is a float as is the value returned. 
18   
19  """ 
20  from Numeric import * 
21   
22     
23 -class ActFunc(object): 
24    """ "virtual base class" for activation functions 
25   
26    """ 
27 -  def __call__(self,input): 
28      return self.Eval(input) 
29   
30   
31 -class Sigmoid(ActFunc): 
32    """ the standard sigmoidal function """ 
33 -  def Eval(self,input): 
34      return 1./(1.+exp(-self.beta*input)) 
35   
36 -  def Deriv(self,input): 
37      val = self.Eval(input) 
38      return self.beta * val * (1. - val) 
39   
40 -  def DerivFromVal(self,val): 
41      return self.beta * val * (1. - val) 
42   
43 -  def __init__(self,beta=1.): 
44      self.beta=beta 
45       
46 -class TanH(ActFunc): 
47    """ the standard hyperbolic tangent function """ 
48 -  def Eval(self,input): 
49      v1 = exp(self.beta*input) 
50      v2 = exp(-self.beta*input) 
51      return (v1 - v2)/(v1 + v2) 
52     
53 -  def Deriv(self,input): 
54      val = self.Eval(input) 
55      return self.beta * (1 - val*val) 
56   
57 -  def DerivFromVal(self,val): 
58      return self.beta * (1 - val*val) 
59   
60 -  def __init__(self,beta=1.): 
61      self.beta = beta 
62