Home | | Artificial Intelligence | Probability and Bayes’ Theorem

Chapter: Artificial Intelligence(AI) : Planning and Machine Learning

Probability and Bayes’ Theorem

In probability theory, Bayes' theorem relates the conditional and marginal probabilities of two random events.

Probability and Bayes’ Theorem

 

In  probability  theory,  Bayes'  theorem  relates the conditional and marginal probabilities of two    random events.

Probability : The Probabilities are numeric values between 0 and 1 (both inclusive) that represent ideal uncertainties (not beliefs).

 

â–  Probability of event A is P(A)


P(A) = 0               indicates total uncertainty in A,

P(A) = 1               indicates total certainty and

0< P(A) < 1 values in between tells degree of uncertainty

Probability Rules :

 

All probabilities are between 0 and 1 inclusive 0 <= P(E) <= 1.

 

The sum of all the probabilities in the sample space is 1.

 

The probability of an event which must occur is 1.

 

The probability of the sample space is 1.

 

The probability of any event which is not in the sample space is zero.

 

The probability of an event not occurring is P(E') = 1 - P(E)

 

Example 1 : A single 6-sided die is rolled.

 

What is the probability of each outcome?

 

What is the probability of rolling an even number?

 

What is the probability of rolling an odd number?

 

The possible outcomes of this experiment are 1, 2, 3, 4, 5, 6.

 

The Probabilities are :

 

P(1)   =       No of ways to roll 1 / total no of sides       = 1/6

P(2)   =       No of ways to roll 2 / total no of sides       = 1/6

P(3)   =       No of ways to roll 3 / total no of sides       = 1/6

P(4)   =       No of ways to roll 4 / total no of sides       = 1/6

P(5)   =       No of ways to roll 5 / total no of sides       = 1/6

P(6)   =       No of ways to roll 6 / total no of sides       = 1/6

P(even)       =       ways to roll even no / total no of sides       = 3/6 = 1/2

P(odd)         =       ways to roll odd  no / total no of sides       = 3/6 = ½

 

Example 2 : Roll two dices

 

Each dice shows one of 6 possible numbers;

 

Total unique rolls is 6 x 6 = 36;

 

List of the joint possibilities for the two dices are:

 


Roll two dices;   

The rolls that add up to 4 are ((1,3), (2,2), (3,1)).

The probability of rolling dices such that  total of 4 is 3/36 = 1/12 and the chance of it being true is (1/12) x 100 = 8.3%.

 

â–   Conditional probability  P(A|B)

A conditional probability is the probability of  an event  given that another event has occurred.

 

Example : Roll two dices.

 

What is the probability that the total of two dice will be greater than 8 given that the first die is a 6 ?

First List of the joint possibilities for the two dices are:


 

There are 6 outcomes for which        the first die is a 6, and of these, there are 4 outcomes that total more than 8 are (6,3;       6,4; 6,5; 6,6).

The probability of a total > 8  given that    first   die is 6 is    therefore 4/6 = 2/3 .


Read as "The probability that the total is > 8 given that die one is 6 is 2/3."

Written as P(A|B) , is the probability of event A given that the event B has occurred.

 

 

â–   Probability of and B is P(A and B)

The    probability that    events A and B both occur.

Note : Two events         are independent if the occurrence of one is unrelated to the probability of the occurrence of the other.

 

‡ If A and B are independent

 

then probability that events A and B both occur is:

 

P(A and B) = P(A) x P(B)

 

ie product of probability of A and probability of B.

 

‡ If A and B are not independent

 

then probability that events A and B both occur is:

 

P(A and B) = P(A) x P(B|A)  where

 

P(B|A) is conditional probability of B given A

 

Example 1: P(A and B) if events A and B are independent

 

Draw a card from a deck , then replace it, draw another card. Find probability that 1st card is Ace of clubs (event A) and 2nd card is any Club (event B).

 

Since there is only one Ace of Clubs, therefore probability P(A) = 1/52.

 

Since there are 13 Clubs, the probability P(B) = 13/52 = 1/4.

 

Therefore, P(A and B) = p(A) x p(B) = 1/52 x 1/4 = 1/208.

 

Example 2: P(A and B) if events A and B are not independent

 

Draw a card from a deck, not replacing it, draw another card. Find probability that both cards are Aces ie the 1st card is Ace (event A) and the 2nd card is also Ace (event B).

 

Since 4 of 52 cards are Aces, therefore probability P(A) = 4/52 = 1/13. 

Of the 51 remaining cards, 3 are aces. so, probability of 2nd


 

â–   Probability of A or B is P(A or B)

 

The probability of either event A or event B occur.

Two events are mutually exclusive if they cannot occur at same time.

 

‡ If A and B are mutually exclusive

 

then probability that events A or B occur is:

P(A or B) = p(A) + p(B)

ie sum of probability of A and probability of B

‡ If A and B are not mutually exclusive then probability that events A and B both occur is:

P(A or B) = P(A) x P(B|A) – P(A and B) where

 

P(A and B) is probability that events A and B both occur while events A and B are independent and P(B|A) is conditional probability of B given A.

 

Example 1: P(A or B) if events A or B are mutually exclusive

 

Rolling a die.

 

Find probability of getting either, event A as 1 or event B as 6? Since it is impossible to get both, the event A as 1 and event B

 

as 6 in same roll, these two events are mutually exclusive. The probability P(A) = P(1) = 1/6 and P(B) = P(6) = 1/6

 

Hence probability of either event A or                                  event B is :

 

P(A or B) =  p(A) + p(B) = 1/6 + 1/6                                  = 1/3

 

Example 2: P(A or B) if events A or B are not mutually exclusive

 

Find probability that a card from a deck    will be either         an  Ace or a          Spade?                          

probability P(A) is  P(Ace) = 4/52 and P(B) is     P(spade) = 13/52.

Only way in a single draw to be Ace and Spade  is Ace of

Spade;        which is  only one, so probability              P(A and  B) is

 

P(Ace and Spade) = 1/52.

 

Therefore, the probability of event A or B is :

 

P(A or B) = P(A) + P(B) – P(A and B)

 

= P(ace) + P(spade) - P(Ace and Spade)

 

= 4/52         + 13/52       - 1/52  = 16/52 = 4/13

 

Summary of symbols & notations


 

Bayes’ Theorem

.

 

Bayesian view of probability is related to degree of belief.

 

It is a measure of the plausibility of an event given incomplete knowledge.

 

 

Bayes' theorem is also known as Bayes' rule or Bayes' law, or called Bayesian reasoning.

 

The probability of an event A conditional on another event B ie P(A|B) is generally different from probability of B conditional on A ie P(B|A).

 

There is a definite relationship between the two, P(A|B) and P(B|A), and Bayes' theorem is the statement of that relationship.

 

Bayes theorem is a way to calculate P(A|B) from a knowledge of P(B|A).

 

Bayes' Theorem is a result that allows new information to be used to update the conditional probability of an event.

 

Bayes' Theorem

 

Let  S  be a sample space.

 

Let  A1, A2, ... , An be a set of mutually exclusive events from S.

 

Let B be any event from the same S, such that P(B) > 0. Then Bayes' Theorem describes following two probabilities :


 

by invoking the fact P(Ak ∩ B) = P(Ak).P(B|Ak) the probability


 

Applying Bayes' Theorem :

 

Bayes' theorem is applied while following conditions exist.

 

the sample space S is partitioned into a set of mutually exclusive events {A1, A2, . . . . . , An }.

 

within S, there exists an event B,  for which P(B) > 0.

 

the goal is to compute a conditional probability of the form :

 

P(Ak|B).

 

you know at least one of the two sets of probabilities described below

 

P(Ak ∩ B) for each Ak

 

P(Ak) and P(B|Ak) for each Ak

 

The Bayes' theorem is best understood through an example below.

 

 

Example 1: Applying Bayes' Theorem

 

 

Problem : Marie's marriage is tomorrow.

 

in recent years, each year it has rained only 5 days. the weatherman has predicted rain for tomorrow.

when it actually rains, the weatherman correctly forecasts rain 90% of the time.

 

when it doesn't rain, the weatherman incorrectly forecasts rain 10% of the time.

The question : What is the probability that it will rain on the day of Marie's wedding?

Solution : The sample space is defined by two mutually exclusive events

"it rains" or "it does not rain". Additionally, a third event occurs when the "weatherman predicts rain".

The events and probabilities are stated below.

 

 

Event A1  : rains on Marie's wedding.

 

Event A2  : does not rain on Marie's wedding

 

Event B   : weatherman predicts rain.

 

P(A1)= 5/365 =0.0136985 [Rains 5 days in a year.]

 

P(A2)= 360/365 = 0.9863014 [Does not rain 360 days in a year.]

 

P(B|A1)= 0.9 [When it rains, the weatherman predicts rain 90% time.]

 

P(B|A2)= 0.1 [When it does not rain, weatherman predicts rain 10% time.]

 

We want to know P(A1|B), the probability that it will rain on the day of Marie's wedding, given a forecast for rain by the weatherman.

 

The answer can be determined from Bayes' theorem, shown below.

 


So, despite the weatherman's prediction, there is a good chance that Marie will not get rain on at her wedding.

 

Thus Bayes theorem is used to calculate conditional probabilities.

 

Example 2: Applying Bayes' Theorem

 

‡  Let  S be a sample space.

 

‡  Let  E1  and E2  be  two mutually exclusive  events forming a partition

 

of the sample space S

 

‡  Let  E be any event of the sample space such that P(E) ≠ 0.


 

Recall from Conditional Probability

 

The notation P(E1 | E) means "the probability of the event E1 given that E has already occurred".

 

‡               The sample space S is described as "the integers 1 to 15" and is partitioned into :

 

E1 = "the integers 1 to 8" and

 

E2 = "the integers 9 to 15".

 

‡ If E is the event "even number" then the probabilities for the situation described by Baye's Theorem can be calculated in two ways, both giving same results.

 


 

Example 3 : Clinic Trial

 

In a clinic, the probability of the patients having HIV virus is 0.15.

 

A blood test done on patients :

 

If patient has virus, then   the test is +ve with probability 0.95.

 

If the patient does not have the virus, then the test is +ve with probability 0.02.

 

Assign labels to events :

 

Given :

 

Find  :

 

If the test is +ve what are the probabilities that the patient

 

i) has the virus         ie P(H|P) ;      ii) does not have virus ie P(¬H|P) ;

 

If the test is -ve  what are the probabilities that the patient

 

iii) has the virus  ie P(H|¬P) ;       iv) does not have virus ie P(¬H|¬P) ;

 

Calculations :

For P(H|P) we can write down Bayes Theorem as

 

P(H|P) = [ P(P|H) P(H) ] / P(P)

 

We know P(P|H) and P(H) but not P(P) which is probability of a +ve result. There are two cases, that a patient could have a +ve result, stated below :

 

Patient has virus and gets a +ve result : H ∩ P

 

Patient does not have virus and gets a +ve result: ¬H ∩ P Find probabilities for the above two cases and then add

 

ie  P(P) = P(H ∩ P) + P(¬H ∩ P).

 

But from the second axiom of probability we have :

 

P(H ∩ P) = P(P|H) P(H) and  P(¬H ∩ P) = P(P|¬H) P(¬H).

 

Therefore putting these we get :

 

P(P) = P(P|H) P(H) + P(P|¬H) P(¬H) = 0.95 × 0.15 + 0.02 × 0.85 = 0.1595

 

Now substitute this into Bayes Theorem and obtain P(H|P)


 


Study Material, Lecturing Notes, Assignment, Reference, Wiki description explanation, brief detail
Artificial Intelligence(AI) : Planning and Machine Learning : Probability and Bayes’ Theorem |


Privacy Policy, Terms and Conditions, DMCA Policy and Compliant

Copyright © 2018-2024 BrainKart.com; All Rights Reserved. Developed by Therithal info, Chennai.