Unformatted text preview: decides to spend a semester abroad at MIT for her Muggle Studies class. Upon
arriving, Hermione befriends the coolest people on campus. The 034 staff, over a festive round of
Butter Beer, then teach Hermione about neural nets. She decides to practice.
Note that the tear off sheets include the neural net tear-off sheet from Quiz 3. X wXA A
-1 Y wYB B
-100 Z wZC wBD wCD D
-1 Hermione creates the above neural net. Each node uses a standard sigmoid function. All threshold units
have fixed weights of 1. They are never updated. 21 Part A: (12 Points)
For each node (A, B, C, and D), using only the weight variables provided in the diagram, d (desired
output), and the outputs of each node (OA, OB, OC, and OD), determine δA, δB, δC, and δD. Part B: (12 Points)
Hermione, as an intellectual exercise, runs the neural net with the following parameters:
WXA = 2
WYA = 2 WYB = 1 WBD = 1.5 WZC = 0.2 WAD = 1 WCD = -1 Determine the output for each node (OA, OB, OC, and OD). Approximate sigmoid(x) very simply by
letting it evaluate to 0 for x < -50 and 1 for x > 50. 22 Part C:...
View Full Document
This document was uploaded on 03/17/2014 for the course EECS 6.034 at MIT.
- Fall '10
- Artificial Intelligence