This preview shows page 1. Sign up to view the full content.
Unformatted text preview: e ID-Tree, beginning with the test Nora has selected from the three tests she
considers, that will correctly classify each Sample 1-8. You can use the grain-size test or any test
of the form Density > a threshold or any test of the form Hardness > a threshold. Select tests that
minimize disorder. Part B3 (4 points) How does you ID-Tree classify Sample A? Sample B? 6 Problem 2: Neural Nets (50 points)
Part A (12 points)
Lyla wants to train a neural net, but she has forgotten what the sigmoid function is. “Oh
well,” she says to herself, “it probably wasn't important anyway.”
She sets up a neural net where the nodes are not sigmoid functions; they are simply
adders. Each node simply adds up its weighted inputs and outputs the sum, so
output = weighted sum of inputs
instead of output= σ(weighted sum of inputs). The derivative of the adder's output with respect to
the weighted sum of the inputs is 1 everywhere.
output weighted sum of inputs...
View Full Document
- Fall '10
- Artificial Intelligence