This preview shows page 1. Sign up to view the full content.
Unformatted text preview: _ wT=_______
If not, why not? 2. Can a perceptron capture inequalities: given two real-valued inputs A and B, can the perceptron
output 1 if A<B and 0 otherwise?
If yes give weights: wA=_______ wB=_______ wT=_______
If not, why not? 3. You wonder about the real-valued function A=B within epsilon E, that is, with three inputs, A, B, and
and E, can a perceptron capture whether |A-B| < E?
If yes give weights: wA=_______ wB=_______ wE=_______ wT=_______
If not, why not? 7 4. The questions in part A have asked for solutions using a threshold decision function. Do the first
round of training for A→B with a sigmoid decision function. Let the learning rate=1.
A B T wA
0 0 -1
0 1 -1
1 0 -1
1 1 -1 wB wT 0 0 ∑
1 y y*
1 The sigmoid decision function: y = 1/(1+e-x): 5. Which of the following is true about a perceptron when we use a sigmoid instead of a threshold?
A. The perceptron can learn XOR
B. The perceptron can no longer learn all linear classification boundaries
C. The perceptron will learn A→B in fewer t...
View Full Document
- Fall '10
- Artificial Intelligence