Chapter 11 Artificial Intelligence

1
Chapter 11Artificial Intelligence

• Introduction to CS
• 1st Semester, 2012 Sanghyun Park

2
Outline

• Introduction
• Understanding Images
• Reasoning
• Artificial Neural Networks
• Genetic Algorithms
• Other Areas of Research

3
Definition

• ________________ (AI) is a branch of science
  which deals with helping machines find solutions
  to complex problems in a more _________ fashion
• This generally involves borrowing characteristics
  from human _________, and applying them as
  _________ in a computer friendly way
• AI is generally associated with ________ science,
  but it has many important links with other fields
  such as math, psychology, cognition, and biology

4
Motivation

• Computers are fundamentally well suited to
  perform __________ tasks efficiently and
  reliably
• Unlike humans, computers have trouble
  understanding ______ situations, and _______ to
  new situations
• AI aims to improve machine behavior in tackling
  such complex tasks
• Much of AI research aims to understand our
  intelligent behavior. Humans have an interesting
  approach to problem-solving, based on _______
  thought, high-level deliberative _________ and
  pattern recognition

5
Eight-Puzzle Machine
The design of eight-puzzle machine provides a
basisfor presenting the topics of the following
two sections
6
Understanding Images (1/2)

• The first intelligent behavior is to understand
  the _____ in order to extract the status of the
  puzzle
• Our machine can detect which tile is in which
  positionby ____-by-____ comparisons
• But it requires a certain degree of ________
  among the style, size and orientation of the
  symbols being read
• Another approach is based on the matching of the
  ________ characteristics
• This method involves two steps
• Extract the features from the image being
  processed
• Compare the features to those of known symbols

7
Understanding Images (2/2)

• The task of understanding general images is
  usually approached as a two-step process image
  processing and image analysis
• Image processing refers to ________
  characteristics of the image
• Edge enhancement
• Region finding
• Smoothing
• Image analysis refers to the process of
  ____________ what these characteristics mean

8
Reasoning

• Once our puzzle-solving machine has deciphered
  the _______ of the tiles, its task becomes that
  of figuring out what ______ are required to solve
  the puzzle
• An approach to this problem is to __________ the
  machine with solutions to all possible
  arrangements of the tiles
• However this approach is not possible when time
  and storage _________ are considered
• Therefore the machine must be programmed to
  perform elementary _________ activities

9
Production Systems

• A large class of reasoning problems have common
  characteristics these common characteristics are
  isolated in a system known as a _________ system
• A production system consists of three main
  components
• A collection of _____ each state is a situation
  that might occur in the application environment
• A collection of _________ a production is an
  operation to shift the system from one state to
  another
• A _____ system it consists of the logic that
  solves the problem of moving the system from the
  _____ state to the goal state

10
State Graph

• An important concept in the development of a
  control system is that of a _____ graph
• A state graph consists of a collection of nodes
  representing the _____ in the system connected by
  arrows representing the __________ that shift the
  system from one state to another
• When viewed in terms of the state graph, the
  problem faced by the control system becomes that
  of finding a sequence of _____ that leads from
  the start state to the ____ state

11
State Graph Example
12
Search Trees

• We have seen that the control systems job
  involves searching the state graph to find a ____
  from the start node to the goal
• One strategy is to construct a ______ tree that
  consists of the part of the state graph that has
  been __________ by the control system
• Let us consider the following start state

13
Sample Search Tree (1/2)
14
Sample Search Tree (2/2)
15
Heuristics (1/2)

• The search tree can become quite _____ if the
  goal is not quickly reached
• One strategy is to change the order in which the
  search tree is constructed rather than building
  it as a _______-first manner, we can pursue the
  more promising paths to greater depths ---
  _____-first construction
• We need a way of identifying which of several
  states appears to be the most ________
• Our approach is to use a _______, which is a
  quantitative value associated with each state
  that estimates the _______ from that state to the
  goal

16
Heuristics (2/2)

• A simple heuristic in the case of the
  eight-puzzle would be to estimate the distance to
  the goal by _______ the number of tiles that are
  out of place
• However this heuristic does not take into account
  how ___ out of position the tiles are
• A better heuristic is to measure the distance
  each tile is from its destination and add these
  values to obtain a single quantity

Heuristic value is __
17
Algorithm for Control SystemUsing Heuristics
18
ExampleBeginning of Heuristic Search
19
ExampleSearch Tree After Two Passes
20
ExampleSearch Tree After Three Passes
21
ExampleComplete Search Tree
22
Artificial Neural Networks

• CPUs that execute sequences of instructions do
  not seem capable of perceiving and reasoning like
  _______
• Many researchers are turning to machines with
  other architectures one of these is the
  artificial neural network
• Artificial neural networks are constructed from
  many processing units, in a manner that models
  networks of _______ in living biological systems

23
A Neuron in a Living Biological System

• The signals transmitted via a cells ____ reflect
  whether the cell is in an ________ or excited
  state
• This state is determined by the combination of
  signals received by the cells _________

24
Activities Within a Processing Unit

• A processing unit is a simple device that mimics
  this basic understanding of the biological
  _______
• It produces an output of 1 or 0, depending on
  whether its ________ input exceeds a given
  ________ value
• This effective input is a weighted sum of the
  actual inputs

25
Weights Within a Processing Unit

• Representation of a processing unit
• The fact that a weight can be positive or
  negative means that the corresponding input can
  have either an inhibiting or _______ effect on
  the receiving unit
• Actual size of the weight controls the _______ of
  effect
• By _________ the values of the weights throughout
  an artificial neural network, we can program the
  network to respond to different inputs in a
  predetermined manner

26
Example Networks
1 if two inputs differ,0 otherwise
1 if both inputs are 1,0 otherwise
27
Genetic Algorithms

• Genetic algorithm applies our understanding of
  natural _______ to the problem-solving task
• This approach is to intermix the ____ performers
  withina collection of proposed solutions to
  obtain another generation of better proposed
  solutions
• By repeating this process, one hopes to simulate
  the __________ process and ultimately obtain
  solution

28
Process in Genetic Algorithms (1/2)

• One finds a way to represent potential solutions
  as ______ of symbols
• A collection of potential solutions is generated
  and tested
• The better solutions from this collection are
  then ______ to form a new generation of potential
  solutions
• At times random __________ may be inserted during
  the crossing process

29
Process in Genetic Algorithms (2/2)
30
Evolutionary Programming

• The goal is to develop programs by allowing them
  to _______ rather than by explicitly writing
  them
• An important step in this setting is to find ways
  in which parts of programs can be ____________ to
  produce meaningful new programs
• The ________ programming paradigm has proved
  useful in this context
• A major problem is to identify the best
  performers out of a group of programs of which
  none seem to be anywhere close to the desired
  product

31
Natural Language Processing (NLP)

• A statement in a natural language can have
  ________ meanings depending on its context
• To unravel the meaning of a statement in a
  natural language therefore requires several
  levels of analysis
• The first level is ________ analysis that
  performs parsing to identify the grammatical role
  of each word
• The next level is _________ analysis that
  identifies the semantic role of each word in the
  statementMary gave John a card John got a
  card from Mary
• A third level is _________ analysis where the
  context of the sentence is brought into the
  understanding processThe bat flew from his hand

32
Information Retrieval And Extraction

• Another area of research in NLP concerns an
  ______ document rather than individual sentences
• Information retrieval refers to the task of
  identifying documents that ______ to the topic at
  hand
• Information extraction refers to the task of
  extracting information from documents so that it
  takes a formthat is useful in other applications
• One such form is known as a _______ that is
  essentially a questionnaire in which specifics
  are recorded
• Another form in which information extractors
  record information is known as _________ net

33
A Semantic Net
34
Robotics

• The goal of the early research in robotics was to
  develop economically viable ________ line robots
  that could increase both productivity and
  consistency
• Today, a major goal of research in robotics is to
  build __________ robots that can maintain their
  balance, walk up stairs, and navigate through
  rough terrain
• Many creative techniques are being applied.One
  is the application of ___________ theories to
  robot development, which generated the field of
  evolutionary robotics

35
Database Systems

• AI techniques are applied to traditional DB
  systems to provide better services. DB techniques
  are applied in AI projects to handle ________
  amounts of knowledge
• One topic is to identify and retrieve information
  that is ______ to a topic rather than merely the
  information that is explicitly requested
• Another topic is the development of data storage
  and retrieval system that can provide information
  that is _______ by the stored data rather than
  merely respond with information that is
  explicitly stored

36
Expert Systems

• Expert systems are software packages designed to
  assist humans in situations where an expert in a
  specific area is required
• These systems simulate the cause-and-effect
  reasoning that experts would accomplish if
  confronted with the same situations
• A major task in constructing an expert system is
  to obtain the required __________ from an expert
• A next task is to _______ the knowledge into a
  format compatible with a software system
• Knowledge base vs. inference engine