SolSet3

SolSet3 - EE 562a 1 Solution Homework Solutions 3 October 1...

This preview shows pages 1–4. Sign up to view the full content.

EE 562a Homework Solutions 3 October 1, 2009 1 1. Solution: (a) In this problem we will use the shorthand notation of m i = the i th component of m y k ij = the ij th element of K y First consider event A : Event A y(u,2) y(u,1) In order to bound the probability with Chebychev’s inequality, we use a little trick: P ( A ) = P { y ( u, 1) < 0 } = P { y ( u, 1) - m 1 < - m 1 } ≤ P {| y ( u, 1) - m 1 | > m 1 } E {| y ( u, 1) - m 1 | 2 } m 2 1 = k 11 m 2 1 = P ( A ) 2 / 10 2 = 0 . 02 . The first inequality is from the bound P { x ( u ) > a } ≤ P {| x ( u ) | > a } . For the event B : Event B 40 -40 y(u,2) y(u,1) y(u,2)=y(u,1)-40 To bound the probability, consider the random variable z ( u ) = y ( u, 1) - y ( u, 2) . The mean and variance of z ( u ) are given by m z = E { z ( u ) } = E { y ( u, 1) - y ( u, 2) } = m 1 - m 2 = 0

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

View Full Document
2 EE 562a Homework Solutions 3 October 1, 2009 V ar { z ( u ) } = E { ( z ( u ) - m z ) 2 } = E { ([ y ( u, 1) - m 1 ] - [ y ( u, 2) - m 2 ]) 2 } = E { ( y ( u, 1) - m 1 ) 2 } - 2 E { ( y ( u, 1) - m 1 )( y ( u, 2) - m 2 ) } + E { ( y ( u, 2) - m 2 ) 2 } = k 11 - 2 k 12 + k 22 = 6 . Then P ( B ) = P { z ( u ) > 40 } ≤ P {| z ( u ) | > 40 } ≤ E {| z ( u ) | 2 } 40 2 = 6 / 1600 = 0 . 00375 Event C corresponds to the outside of a circle of radius 10, centered at m y since k y ( u ) - m y k > 10 ⇐⇒ k y ( u ) - m y k 2 > 100 , k y ( u ) - m y k 2 = ( y ( u, 1) - m 1 ) 2 + ( y ( u, 2) - m 2 ) 2 Event C 10 10 y(u,2) y(u,1) Chebychev’s inequality applies directly in this case P ( C ) = P {k y ( u ) - m y k > 10 } ≤ E {k y ( u ) - m y k 2 } 10 2 = Tr ( K y ) 10 2 = P ( C ) 4 / 100 = 0 . 04 Remark : The easy way to determine the region of the event is to draw the boundary first - that is turn the inequality into an equality and plot it in the plane. This gives a line, circle, etc., so to find out on which side of the boundary the event lies test a simple point (the origin is the easiest if it is not on the boundary). (b) We have already evaluated the mean and variance of the random variables in this prob- lem. Since y ( u, 1) is a Gaussian random variable, and since z ( u ) is Gaussian because it is generated by a linear operation on y ( u ), the answers here are simply evaluations of Gaussian integrals. First consider event A : P ( A ) = P { y ( u, 1) < 0 } = Z 0 -∞ p y ( u, 1) ( x ) dx = Z 0 -∞ exp ( - ( x - m 1 ) 2 / 2 k 11 ) 2 πk 11 dx Putting this integral in standard form by substituting z = ( x - m 1 ) / k 11 gives P ( A ) = Z - m 1 / k 11 -∞ exp( - z 2 / 2) 2 π dz = Z - 10 / 2 -∞ exp( - z 2 / 2) 2 π dz exp( - 7 . 07 2 / 2) 7 . 07 2 π 7 . 9 × 10 - 13
EE 562a Homework Solutions 3 October 1, 2009 3 Coming up with a numerical answer may have been difficult for you here. Of course computers are always available. But computers rely on numerical algorithms and some very good ones are available in Abramowitz and Stegun (see hint). I used the symmetry of the probability density of an elementary Gaussian random variable and relations 26.2.12 and 26.2.3 above to come up with an answer that provides accuracy to less than 2 percent error. Note the significant difference in Chebychev’s error probability bound of part (a) and the result when the variables are known to be Gaussian. Similarly for B , using the derived statistics for the random variable z ( u ) gives P ( B ) = P { z ( u ) > 40 } = Z 40 p z ( u ) ( x ) dx = Z 40 exp( - x 2 / 2 · 6) 2 π · 6 dx and setting z = x/ 6 gives P ( B ) = Z 40 / 6 exp( - x 2 / 2) 2 π dx exp( - 16 . 33 2 / 2) 16 . 33 2 π 3 . 0 × 10 - 60 That is a small number!

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

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

{[ snackBarMessage ]}

What students are saying

• As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

Kiran Temple University Fox School of Business ‘17, Course Hero Intern

• I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

Dana University of Pennsylvania ‘17, Course Hero Intern

• The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

Jill Tulane University ‘16, Course Hero Intern