quiz3_2009

The perceptron will learn ab in fewer training steps

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: raining steps D. The perceptron will learn A→B in more training steps E. None of these is true 8 Part B (20 points) Given this three-node neural network, and the training data on the right x y w11 w12 w21 w1T w31 -1 w v -1 3T w32 w22 w2T -1 1. Indicate which of the following sets of weights will separate the dots from the diamonds by circling their letters. NOTE: more than one may work! w 11 A B C D E F w 12 3 -2 1 4 2 4 w 1T 2 -2 1 4 -2 4 w 21 1 -1 0.5 2 1 2 w 22 4 3 1 6 2 6 5 3 1 6 -2 6 w 2T w 31 w 32 w 3T -2 -100 -100 -150 -1.5 128 128 173 -0.5 97 97 128 -3 96 -95 -52 1 100 100 50 -3 -101 102 148 2. For (at least) one of the sets of weights you chose above, write the simplest mathematical representation for it using + - * /, inequalities, and/or boolean operations. Which line? mathematical expression: 3. When training the three-node neural network at the top of the page using back-propagation, if the current weights are the ones given in choice A, then, for the training example x=0 y=0, if y*=1, what is δ1? See the tear-off sheet for notes on back-propagation. You...
View Full Document

This document was uploaded on 03/17/2014 for the course EECS 6.034 at MIT.

Ask a homework question - tutors are online