15-780: Graduate AI
Homework Assignment #3
Out: March 23, 2015
Due: April 6, 2015 5 PM
Collaboration Policy: You may discuss the problems with others, but you must write all
code and your writeup independently.
Turning In: Please email your assignment by
3/23/15
Bayes Networks (cont) and
POMDPs (just a start)
Manuela M. Veloso
Carnegie Mellon University
Computer Science Department
Thanks to past instructors
15-780 Graduate AI Spring 2015
Russell & Norvig: chapter 17
Bayes Net Example CPTs
1
3/23/15
Bayesi
3/18/15
Bayes Networks: Representation
and Inference
Manuela M. Veloso
Carnegie Mellon University
Computer Science Department
Thanks to past instructors
15-780 Graduate AI Spring 2015
Readings: Russell & Norvig: chapter 14
Where are We
States, reward, act
CMU 15-780
Emma Brunskill (THIS TIME)
Manuela Veloso
What is probability?
Frequentists
Bayesians
Frequency of Event
Degree of Belief
2
Axioms of Probability
Let A be a proposition about the world
P(A) = probability proposition A is true
0 <= P(A) <= 1
P
15-780
Spring 2015
HMMs
Emma Brunskill (this time)
Manuela Veloso
With thanks to Dan Klein and Andrew Moore for slides and
slide inspirations
1
HMM Definition
An HMM is defined by:
Initial distribution: P(X0)
Transitions:
Observations:
X
X
X
X
1
2
3
4
Reinforcement Learning+
Emma Brunskill (today)
Manuela Veloso
1
Q-Learning Recap
2
3
The Speed of Learning and
Speeding Learning
4
5
ObjecJves for an RL Algorithm
AsymptoJc guarantees
In limit converge to
Uninformed Search
Manuela Veloso
Carnegie Mellon University
http:/www.cs.cmu.edu/~15780
15-780 Spring 2015
Russell & Norvig, Chapters 1, 2, Sections 3.1-3.4.
(Thanks to past instructors for all lecture slides.)
Search Problem Representation:
Just a Graph
15-780
Spring 2015
HMMs
Emma Brunskill (this time)
Manuela Veloso
With thanks to Dan Klein and Andrew Moore for slides and
slide inspirations
1
Probabilistic Inference
Compute probability of a query variable(s)
given some evidence
2
Reasoning over Time
15-780
Click to edit Master 4tle style
MDPs
Click to edit Master sub4tle style
Emma Brunskill (this 4me)
Manuela Veloso
2/18/15
2/18/15
1
1
Projects
If you already emailed us, great!
If you havent, rememb
CMU 15-780
Informed Search
Teachers:
Emma Brunskill (this time)
Manuela Veloso
With thanks to Ariel
Procaccia and other prior
instructions for slides
Problem Solving
Given
o
o
o
An initial state
A set of actions
A goal statement
Find a plan, a sequence
(L) /\ (:3 {I
1:
f if? 5 z" ' l' r) A
i if} i; 4'7? 5 5? :i'z.) rm ,5 v" V
i
If,
f (5 f (if
[m
71} Kim» /V 9"?
[A
k/ (73 f
w" ,1 m I
V I0 J «23 :3 37;: m is»: (3* f W i2, 5/ {7 f z E!
w
53y fame/Om t/griééx/é? ,
573m; iv: 1? «fr 3:0 mm r [n /'>/C>:
15-780 Grad AI
CSPs and Local Search
Spring 2015
Manuela Veloso
Emma Brunskill
Chapters 5 and 4, Russell and Norvig
Thanks to all past AI instructors
Carnegie Mellon
Outline
Examples and definitions
Standard search
Improvements
Backtracking
Forward c
15-780: Graduate AI
Homework Assignment #1 Solutions
Out: January 22, 2015
Due: February 4, 2015 5 PM
Collaboration Policy: You may discuss the problems with others, but you must write all
code and your writeup independently.
Turning In: Please email your
15-780: Graduate AI
Homework Assignment #2 Solutions
Out: February 12, 2015
Due: February 25, 2015
Collaboration Policy: You may discuss the problems with others, but you must write
all code and your writeup independently.
Turning In: Please email your as
15-780: Graduate AI
Homework Assignment #2
Out: February 12, 2015
Due: 5 PM February 25, 2015
Collaboration Policy: You may discuss the problems with others, but you must write
all code and your writeup independently.
Turning In: Please email your assignm
15-780 Convex Optimization
Shayan Doroudi
January 29, 2015
Thanks to Zico Kolter for the original slides!
1
Introduction to Mathematical Optimization
Casting AI problems as optimization / mathematical
programming problems has been one of the primary tren
Grad AI
Constraint Satisfaction Problems
Spring 2015
Manuela Veloso
Emma Brunskill
Chapter 5, Russell and Norvig
Thanks to all past AI instructors
Carnegie Mellon
Outline
Examples and definitions
Standard search
Improvements
Backtracking
Forward chec
2/9/15
Classical Planning
Manuela M. Veloso
Carnegie Mellon University
Computer Science Department
15-780 Graduate AI Spring 2015
Readings:
Chapter 10, Russell & Norvig
Integrating planning and learning: The Prodigy architecture, (Sections 1 and 2)
M M.
2/25/15
Reinforcement Learning:
Q-Learning
Manuela M. Veloso (this time)
Emma Brunskill
15-780 Graduate AI Spring 2015
Readings:
Russell & Norvig: chapter 21
Machine Learning, Tom Mitchell, chapter 13
Reinforcement Learning: An Introduction, R. Sutton
15-780: Spring 2015 Graduate AI
Introduction
Instructors:
Emma Brunskill and Manuela Veloso
TAs:
Shayan Doroudi and Vittorio Perera
http:/www.cs.cmu.edu/~15780
Carnegie Mellon University
Evaluation
n
n
n
n
n
n
n
4 Homeworks
Test 1
Final Project
Test 2
40%