Go backward to 2 Decision Tree Evaluation
Go up to Top

# 3 Neural Network

Suppose that a neural network learner uses the network corresponding to the rule:
```predicted_prop(Obj,d,V) <-
prop(Obj,a,I_1)&
prop(Obj,b,I_2)&
prop(Obj,c,I_3)&
V is f(w_0 + w_1*I_1 + w_2*I_2 + w_3*I_3).
```
where f is the sigmoid function.
f(x)= 1/(1+e-x)
(The only property of f you need for this exam is that f(x) > 0.5 if and only if x>0.)

Suppose that, after learning, the parameters had the following weights:

 w0 -3 w1 2 w2 2 w3 4
Suppose the neural network classifies as true any example where the predicted value for d is greater than 0.5
1. How is example e1 classified, where e1 is defined by:
```prop(e_1,a,1).
prop(e_1,b,1).
prop(e_1,c,0).
```
2. Give an example that is classified differently by the neural network and the decision tree
if(b=1,if(a=1,true,false),if(c=1,false,true)).
(which is equivalent to the example decision tree given in the previous problem).
3. Draw a decision tree that represents the same Boolean function as that represented by the neural network.
• Solution to part (a)
• Solution to part (b)
• Solution to part (c)

• Computational Intelligence online material, ©David Poole, Alan Mackworth and Randy Goebel, 1998