barkerma-hw6 - hw6-inv.txt Enter name of test data:...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Michael Barker CSA 486 HW 6 1.) P ( 1 ) = P( 1 | F ) P( F ) + P( 1 | R ) P( R ) = .25*.75 + .5*.25 = .3125 P( F | 1 ) = P( 1 | F ) P( F ) / P( 1 ) = .25*.75/.3125 = .6 P( F | 2(1s) ) = P( 2(1s) | F ) P( F ) / P ( 2(1s) ) = (.25 2 ).75/(.3125 2 ) = .48 P( F | 3(1s) ) = P( 3(1s) | F ) P( F ) / P ( 3(1s) ) = (.25 3 ).75/.3125 3 = .384 P( F | 4(1s) ) = P( 4(1s) | F ) P( F ) / P ( 4(1s) ) = (.25 4 ).75/.3125 4 = .3072 P ( 1 | 4(1s) ) = P( 1 | F ) P( F | 4(1s) ) + P ( 1 | R ) P( R | 4(1s) ) = .25*.3072 + .5*(1-.3072) = .4232 P ( 4 | 4(1s) ) = P( 4 | F ) P( F | 4(1s) ) + P ( 4 | R ) P( R | 4(1s) ) = .25*.3072 + 0*(1-.3072) = .0768 2.) Learning Rate Epochs Training Set Performance W1 W2 W3 0.005 2000 1.0 0.29527634120 3539646193173 -0.82867258847 0.005 20 0.666667 0.05182584271 0.07101642852 -0.01770250402 0.99 2000 0.666667 0.53295513958 0.48744232563 -0.64517500926 0.000001 2000 0.666667 0.00132620159 0.00176916852 -0.00022544290 3.) Enter name of training data:
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 2
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: hw6-inv.txt Enter name of test data: hw6-inv.txt Number of epochs: 2000 Learning rate: .005 Perceptron weights: 0.2950693419412812 0.3538241330553852 -0.8287566968759267 Training performance: 1.000000 The weights and performance are the same. The results are the same due to the line that seperates the data points being the same. Since the location of the points was not changed and all of the values were inverted the line that divides the data points remains the same. 4.) The training performance is 0.8. The best possible performance that one could hope for is 0.09. Due to perceptrons being linear it is impossible to have a 1.0 performance rate for the training data and in this case the best that can happen is to correctly match 9 points and not match the new point located at (-1, 1). 5.) BUPA Performance: 0.6231884057971014 Spam Performance: 0.830471636600739...
View Full Document

Page1 / 2

barkerma-hw6 - hw6-inv.txt Enter name of test data:...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online