Download Probability Models - Brown Computer Science

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Probability Models
Weather example:
Venn diagram for all
combinations of 3
binary (true/false) events.
Ø
Ø
Ø
●
●
●
Raining or not
Sunny or not
Hot or not
Sample spaces: When a random event happens, what is the
set of all possible outcomes? May be discrete or continuous.
Conditioning: Suppose I observe some data. How does
my probability model change?
Independence: Is there any relationship between pairs of
variables in my model? Would data provide knowledge?
Defining a Probabilistic Model
flip a coin,
roll a die,
receive an email,
take a picture, …
The Axioms of Probability
event
A
Ø
Ø
Ø
Ø
event
B
Valid probabilities defined by any function mapping subsets of
to [0,1] that satisfies these axioms (assumptions)
The nonnegativity and additivity axioms are fundamental to
our intuitive understanding of probability and uncertainty
Unit normalization is just a convention, and other options
could also be used (e.g., probability between 0% and 100%)
The additivity axiom guarantees that the probabilities of any
finite set of disjoint events are additive (induction)
Conditional Probability
Bayes’ Rule
Independence of Two Events
Computing a PMF
Several Random Variables
y
z
May compute marginal of any subset of variables,
possibly conditioned on values of any other variables.
Expected Values of Functions
Ø
Consider a non-random (deterministic) function of a random variable:
Ø
What is the expected value of random variable Y?
Ø
Correct approach #1:
Ø
Correct approach #2:
Ø
Incorrect approach:
(except in
special cases)
Linearity of Expectation
Ø
Consider a linear function:
Ø
Ø
Example: Change of units (temperature, length, mass, currency, …)
Ø
Ø
In this special case, mean of Y is the linear function applied to E[X]:
Expectation of Multiple Variables
Ø
The expectation or expected value of a function of two discrete variables:
Ø
A similar formula applies to functions of 3 or more variables
Ø
Expectations of sums of functions are sums of expectations:
Ø
Ø
This is always true, whether or not X and Y are independent
Specializing to linear functions, this implies that:
Variance and Standard Deviation
Ø
Ø
The variance is the expected squared deviation of a random variable
from its mean (these definitions are equivalent):
By definition, the standard deviation is the square root of the variance:
Variance and Moments
Ø
The variance is the expected squared deviation of a random variable
from its mean (these definitions are equivalent):
Terminology: Moments of random variables
first moment or mean of X
second moment of X
pth moment of X
Continuous Random Variables
CDF: cumulative
distribution function
PDF: probability
density function
Model processes or
data which are encoded
as real numbers:
temperature,
commodity price,
DNA expression level,
light on camera sensor,
…
Continuous Random Variables
Ø
Ø
Ø
For any discrete random variable, the CDF is
discontinuous and piecewise constant
If the CDF is monotonically increasing and
continuous, have a continuous random variable:
The probability that continuous random variable
X lies in the interval (x1,x2] is then
1
0
Probability Density Function (PDF)
Ø
Ø
If the CDF is differentiable, its first derivative is
called the probability density function (PDF):
1
By the fundamental theorem of calculus:
0
Ø
For any valid PDF:
0
Expectations of Continuous Variables
Ø
The expectation or expected value of a continuous random variable is:
Ø
The expected value of a function of a continuous random variable:
Ø
The variance of a continuous random variable:
Ø
Intuition: Create a discrete variable by quantizing X, and compute
discrete expectation. As number of discrete values grows, sum approaches
integral.
Joint Probability Distributions
Marginal Distributions
Example: Uniform Distributions
Independence
Distributions
●
Bernoulli Distribution
●
Gaussian Distribution
●
Exponential Distribution
●
Geometric Distribution
●
Uniform Distribution
●
Formulas will be provided, but you should be familiar with their
properties
Things to Prepare
●
Lecture Notes
●
Homework Questions
●
(Textbook Problems)
●
(Recitation Material)