Download presentation

Presentation is loading. Please wait.

Published byEric Farley Modified over 4 years ago

1
Defender Acts 1st A1A1 A2A2 D1D1 C 11 C 12 D2D2 C 21 C 22 where C i,j =Cost to defender from play (A j |D i ) A1A1 A2A2 D1D1 1 12 D2D2 21 22 Random Cost Matrix Expected Cost Matrix where i,j =E[C i,j ]

2
Minimax Strategy A1A1 A2A2 Max j C i,j D1D1 C 11 C 12 C*1C*1 D2D2 C 21 C 22 C* 2 A1A1 A2A2 Max j i,j D1D1 1 12 * 1 D2D2 21 22 * 2 Random Cost Matrix Expected Cost Matrix

3
Minimax Strategy A1A1 A2A2 Max j C i,j D1D1 N(6,10)N(6,2) C*1C*1 D2D2 N(6.5,3.5)N(7.5,3) C* 2 A1A1 A2A2 Max j i,j D1D1 * 1 D2D2 * 2 Random Cost Matrix Expected Cost Matrix Example: C i,j =N( i,j, i,j )

4
Minimax Strategy A1A1 A2A2 Max j C i,j D1D1 N(6,10)N(6,2) C*1C*1 D2D2 N(6.5,3.5)N(7.5,3) C* 2 A1A1 A2A2 Max j i,j D1D1 * 1 D2D2 * 2 Random Cost Matrix Expected Cost Matrix Example: C i,j =N( i,j, i,j ) Which action should Defender take? D*=argmin i max j E[C i,j ] =argmin i * i

5
Banks and Anderson Strategy #1 Choose D 1, but rather close to indifferent A1A1 A2A2 Max j C i,j D1D1 N(6,10)N(6,2) C*1C*1 D2D2 N(6.5,3.5)N(7.5,3) C* 2 D*=argmax i P(C* i < min k C* k )

6
Banks and Anderson Strategy #2 From this, choose D 2 Score(i)=min k {C* k } / C* i Score(i) 2 (0,1] – Larger is better E[Score(1)]=0.815 E[Score(2)]=0.822 A1A1 A2A2 Max j C i,j D1D1 N(6,10)N(6,2) C*1C*1 D2D2 N(6.5,3.5)N(7.5,3) C* 2 D*=argmax i E[Score(i)]

7
An Alternative Approach where m* i =E[C* i ]=E[max j C i,j ] A1A1 A2A2 Max j C i,j D1D1 N(6,10)N(6,2) C*1C*1 D2D2 N(6.5,3.5)N(7.5,3) C* 2 Choose D 2, since worst case has lower expected cost D*=argmin i E[max j C i,j ]

Similar presentations

Presentation is loading. Please wait....

OK

Arithmetic and Geometric Means

Arithmetic and Geometric Means

© 2018 SlidePlayer.com Inc.

All rights reserved.

To make this website work, we log user data and share it with processors. To use this website, you must agree to our Privacy Policy, including cookie policy.

Ads by Google