ECS 170: Problem Set 2 Solution
January 26, 2017
Note that your solutions may be correct but not identical to our
solutions. We will keep this in mind while grading.
1. (2) (R&N 3.18) Describe a search space in which iterative deepening search
performs mu

Main
Page
Classes
Files
Project 1: A* On Terrain Maps
Introduction
You are lost but fortunately you have an A* framework which you can use to find your way home.
Important: No two adjacent nodes are at height zero. This is important for the division cost

ECS 170: Problem Set 1
January 11, 2017
Your answers should be succinct - our solutions for each problem
are no more than a couple sentences.
Your submission should be a PDF. We make no guarantees we will
grade submissions in other formats.
1. What is the

ECS 170: Problem Set 3
January 25, 2017
Your answers should be succinct - our solutions for each written
problem are no more than a couple sentences.
For this homework you will need to modify a game trees. Because
you are submitting this electronically, w

ECS 170: Problem Set 4
February 8, 2017
Your answers should be succinct.
1. Consider the Bayesian network in Figure 1.
A
B
C
D
Figure 1: Network for problem 1.
(a) How many numbers need to be stored in the probability tables to
represent this model? Show

ECS 170: Problem Set 5
February 25, 2017
Your answers should be succinct - our solutions for each written
problem are no more than a couple sentences.
1. What condition must hold on the training data such that the perceptron
training algorithm can learn a

ECS 170 Assignment Project 1 Part 4:
Kevin Largo & Tiffany Lee
Admissible Heuristic & Modifying A*: We ran out of time, but we would have
changed our implementation to use HashMaps instead of an Array List lik

Tiffany Lee
ECS 170 Homework 1
1/12/2016
1) BFS: uses a queue, DFS: uses a stack, Dijkstra: uses a priority queue.
2) The graph has edges with negative weight.
3) The heuristic given to the A* Algorithm does no

ECS 170 Homework 5
Tiffany Lee
1. (4) Consider a cumulative discount reward (the objective function of Qlearning)
with a = 0 and = 1. What type of behavior would these
reward functions encourage?
If = 0

ECS 170 Homework 4
Tiffany Lee
1. (2) You are experimenting with a pair of Bayesian networks. While the
topology of the networks are different, you notice that for all queries the
two networks output the same

Coding Conventions
Kernel Style
Tab stops 8 characters
Bracket on next line for function
Other brackets on same line
Switch statement aligned with block
Kernighan & Ritchie Style
Bracket on next line for function
Other brackets on same line
Closing

Types of Relationships
BN Rational Reasoning! Networks
Should Capture Our Intuition of
What type
function does
this look like?
15
P increase 2.5 fold
S increases > 2. 5 fold
Why?
Graphs Captures Intuition
18
BN for Bulgaria
How To Calculate Queries
- Basi

Some Logistical
Observations
Alpha and Beta are lower and upper bounds on
the Nash equilibrium.
Max fills in alpha, Min fills in beta.
Why?
Max and min will eventually converge and we get
the Nash equilibrium of the game.
An observation of a payoff c

Main
Page
Classes
Files
Project 1: Connect Four
Due 02/27/17
Introduction
In this assignment, you'll write a computer program to play the classic game Connect Four. This is a great
way to practice the minimax and heuristic algorithms we've covered in clas

How Neurons Process
Information
Lets say we have an OR Percepton
w0 = 1, w1=2, w2=2. w0 called bias and w0
threshold
Calculations
x0
x1
x2
t
o
1
-1
-1
0
1-2-2 <= 0
1
-1
+1
1
1 - 2 +2 > 0
1
+1
-1
1
1+22 > 0
1
+1
+1
1
1+2+2 > 0
31
How Networks of Percept

ECS170 Introduction to A.I.
Course Overview Lecture
Ask Lots of Questions!
What is A.I.?
Course Overview
Course Logistics
Good news and bad news
Lets begin!
Who can tell me some well known AI applications?
1
Well Known AI Applications
Games:
Check

Problem Set 1 Solutions
January 18, 2017
1. (2) What is the difference among BFS, DFS, and uniform-cost search
(Dijkstras algorithm) with respect to their implementations in the generic
tree search algorithm?
From a high level view, the three algorithms o

ECS 170: Homework 2
January 19, 2017
Your answers should be succinct - our solutions for each problem
are no more than a couple sentences.
Your submission should be a PDF. We make no guarantees we will
grade submissions in other formats.
1. (R&N 3.18) Des

ECS 170: Problem Set 3 Solution
February 1, 2017
Note that your solutions may be correct but not identical to our
solutions. We will keep this in mind while grading.
1. Minimax assumes both players are rational. However, suppose one player
is not rational

ECS 170: Problem Set 4 Solution
February 12, 2017
1. (4) Consider the Bayesian network in Figure 1.
A
B
C
D
Figure 1: Network for problem 1.
(a) (2) How many numbers need to be stored in the probability tables to
represent this model? Show your work by li

Faculty Candidate Talk
Title: Balancing risk and performance in robot planning algorithms
When: Tuesday, February 7th at 3:10pm Where: 1131 Kemper Hall
Abstract: Robotics is experiencing a period of explosive growth in academia and industry. As
robots ar

The Story So Far
1. Formulating problems as searching trees/
graphs
2. Briefly cover basics of uninformed search
3. Why we need A*
4. Then move onto Informed search (A*) Algorithm
5. Inventing Admissible Heuristics
6. Performance analysis and data structu

Key Question in
Assignment
rd
3
Explaining how the network is predicting so well
Groups of two. How can we do this?
70
The
Challenges With
Applying
Machine Learning?
71
Assignment #1
Two group members getting different grades
Part 4 grading issue
72
A

Approximate Inference With
Random Walks
Worse case complexity for a query?
Draw the graph.
Think of the SPAM graphical model.
What can we do?
Simplify with the Markov blanket
Calculate approximation
Why is an approximate answer sometimes good
enoug

We Can Query Any Set of Propositions
Given Any Other Set (Evidence)
Belief networks are a concise representation of the joint probability
distribution. A simple abstraction of the world
P(B)=0.001
Burglary
P(E)=0.002
Specifying the
joint distribution
requ

Three Core Parts of AI
Course
Search, Reasoning and Learning
Only search module considers adversarial
environment.
What would it mean for searching in an
adversarial environment for first assignment?
Give me some examples if reasoning/learning in
adve

Modules D) and E):
Machine Learning
So far to get our algorithms to do anything
intelligent, we had to do some work.
Intelligent search: A* ?
Play connect-4: Minimax
Reason what actions to perform: Belief networks
So the algorithms found optimal path

This Where it Gets A Bit
Mind Bending
With mini-max and payoffs we calculate the Nash
equilibrium as it assumes are opponent is behaving
rationally.
But with evaluation functions whats going on?
31
Mind Bending (Like The
Matrix)
With mini-max and payof