ECE 302: Lecture 2.6 Bayes Theorem and Total Probability
Prof Stanley Chan
School of Electrical and Computer Engineering Purdue University
©Stanley Chan 2020. All Rights Reserved. 1 / 1 Outline
2.1 Set theory 2.2 Probability space 2.3 Axioms of probability 2.4 Conditional probability 2.5 Independence 2.6 Bayes theorem 2.6.1 Prisoner’s Dilemma 2.6.2 Bayes Theorem and Law of Total Probability 2.6.3 Examples
©Stanley Chan 2020. All Rights Reserved. 2 / 1 Prisoner’s Dilemma
Three Prisoners: A, B, C. The King decides to release 2 and sentence 1. You were A. Your chance of release is 2/3. Suppose you know the guard well. You can ask him about which of B or C will be released. But if you find out B (or C) is released, your chance becomes 1/2. How come!!
©Stanley Chan 2020. All Rights Reserved. 3 / 1 Conditional Probability
XA = sentence A, XB = sentence B, XC = sentence C. GB = guard says B is released.
Let us find the probabilities: P[XA], P[XB ], P[XC ]
©Stanley Chan 2020. All Rights Reserved. 4 / 1 Conditional Probability
XA = sentence A, XB = sentence B, XC = sentence C. GB = guard says B is released.
Let us find the probabilities: P[GB | XA], P[GB | XB ], P[GB | XC ]
©Stanley Chan 2020. All Rights Reserved. 5 / 1 Stuck...
P[XA | GB ] =
©Stanley Chan 2020. All Rights Reserved. 6 / 1 Outline
2.1 Set theory 2.2 Probability space 2.3 Axioms of probability 2.4 Conditional probability 2.5 Independence 2.6 Bayes theorem 2.6.1 Prisoner’s Dilemma 2.6.2 Bayes Theorem and Law of Total Probability 2.6.3 Examples
©Stanley Chan 2020. All Rights Reserved. 7 / 1 Bayes Theorem
Theorem (Bayes Theorem) For any two events A and B such that P[A] > 0 and P[B] > 0, it holds that P[B | A] P[A] P[A | B] = . P[B]
©Stanley Chan 2020. All Rights Reserved. 8 / 1 Law of Total Probability
Theorem (Law of Total Probability)
Let {A1, A2,..., An} be a partition of Ω, i.e., A1,..., An are disjoint and Ω = A1 ∪ A2 ∪ ... ∪ An. Then, for any B ⊆ Ω,
n X P[B] = P[B | Ai ] P[Ai ]. i=1
©Stanley Chan 2020. All Rights Reserved. 9 / 1 Law of Total Probability
continue...
©Stanley Chan 2020. All Rights Reserved. 10 / 1 Law of Total Probability
Figure: Law of total probability decomposes the probability P[B] into multiple conditional probabilities P[B | Ai ]. The probability of obtaining each P[B | Ai ] is P[Ai ].
©Stanley Chan 2020. All Rights Reserved. 11 / 1 Outline
2.1 Set theory 2.2 Probability space 2.3 Axioms of probability 2.4 Conditional probability 2.5 Independence 2.6 Bayes theorem 2.6.1 Prisoner’s Dilemma 2.6.2 Bayes Theorem and Law of Total Probability 2.6.3 Examples
©Stanley Chan 2020. All Rights Reserved. 12 / 1 Example 1: Tennis Tournament
Example. Consider a tennis tournament. Your probability of winning the game is 1 0.3 against of the players (Event A). 2 1 0.4 against of the players (Event B). 4 1 0.5 against of the players (Event C). 4 What is the probability of winning the game?
©Stanley Chan 2020. All Rights Reserved. 13 / 1 Example 1
continue...
©Stanley Chan 2020. All Rights Reserved. 14 / 1 Example 2: Communication Channel
Example. Consider a communication channel shown below. The probability of sending a 1 is p and the probability of sending a 0 is 1 − p. Given that 1 is sent, the probability of receiving 1 is 1 − η. Given that 0 is sent, the probability of receiving 0 is 1 − ε.
Find P[Receive 1] Find P[Send 1] Find P[Receive 1 | Send 1] Find P[Send 1 | Receive 1] ©Stanley Chan 2020. All Rights Reserved. 15 / 1 Example 2
continue...
©Stanley Chan 2020. All Rights Reserved. 16 / 1 Example 2
continue...
©Stanley Chan 2020. All Rights Reserved. 17 / 1 Prisoner’s Dilemma (continue)
P[XA | GB ] =
©Stanley Chan 2020. All Rights Reserved. 18 / 1 Questions?
©Stanley Chan 2020. All Rights Reserved. 19 / 1