PowerPoint 演示文稿

PART IV:
The Potential
of
Algorithmic Machines
Artificial
Intelligence.
 Theory of Computation.
Ch. 10 Artificial Intelligence







Some philosophical issues.
Image analysis.
Reasoning.
Control system activities.
Using Heuristics.
Artificial neural networks.
Applications of AI.
Some Philosophical Issues

Machines Vs. humans.
 Performance Vs. simulation.
 Intelligence as an interior characteristic Turing test and program DOCTOR (ELIZA).
 How to create an intelligent machine?
10.1 Intelligence and Machines
Turing Test : 1950, Alan Turing proposed a test
to evaluate the intelligent behavior of a
machine.
Figure 10.1: Our puzzle-solving
machine
An Intelligent puzzle-solving
machine

This machine takes the form of a metal box
equipped with a gripper, a video camera,
and a finger with a rubber end so that it
does not slip when pushing something.
 Actions:
1.
Turn on the machine.
2.
Place the puzzle.
3.
The finger pushes the tiles back to the
original order.
4.
Turn off the machine.
10.2 Image Analysis

The first intelligent behavior required by
the puzzle-solving machine is the extraction
of information through a visual medium.
 Perceive ability - determine the current
status of the puzzle.
 Optical character readers.
 Character recognition based on matching
the geometric characteristics.
Figure 10.2: The eight-puzzle in
its solved
configuration
Figure 10.3: A small portion of
the eight-puzzle’s state graph
10.3 Reasoning

Is possible to develop proper programs
targeted to all possible initial configurations
(in total 181,440 of them)?
 Develop a program which can solve the
problem itself - the ability to make
decisions, draw conclusions, and in short,
perform elementary reasoning activities.
Reasoning

A production system consists of three main
components:
 1. A collection of states - start/goal states.
 2. A collection of productions (rules).
 3. A control system - which consists of the
logic that solves the problem of moving
from the start state to the goal state.
 State graph - conceptualizing all states,
rules, and preconditions in a production
system.
Reasoning
Socrates is a man.
All men are humans.
All humans are mortal.
Start state
Goal state
Socrates is a man.
All men are humans.
All humans are mortal.
Socrates is a human.
Socrates is a man.
All men are humans.
All humans are mortal.
Socrates is a human.
Socrates is mortal.
Figure 10.4: Deductive reasoning
in the context of a production
system
Control System Activities

A state-graph traversal problem.
 Search tree.
 How to build a search tree?
 It is impractical to develop a full search
tree for a complex problem.
 Using depth-first construction instead of
breadth-first manner.
 Avoiding redundancy.
Figure 10.5: An unsolved eightpuzzle
Figure 10.6: A sample search
tree (continued)
Figure 10.6:
A sample
search tree
(continued)
Figure 10.6:
A sample
search tree
(continued)
Figure
10.6:
A
sample
search
tree
Figure 10.7: Productions stacked
for later
execution
Figure 10.8: An unsolved eightpuzzle
Figure 10.9: An algorithm for a
control system
using
heuristics
Using Heuristics

Heuristics - the use of intuition, a rule of
thumb which may lead to a correct direction
but offer no assurance on it.
 How to develop a heuristic - first develop a
quantitative measure by which a program
can determine which of several states is
considered closest to the goal (cost
function).
Figure 10.10: The beginning of
our heuristic
search
Figure 10.11: The search tree
after two passes
Figure
10.12:
The
search
tree after
three
passes
Figure
10.13:
The
complet
e search
tree
formed
by our
heuristic
system
10.4 Artificial Neural Networks

Neural networks - model networks of
neurons in living biological systems.
Compute effective
inputs
I1W1+…+InWn
Threshold
value
Output
0 or 1
Figure 10.14: A neuron in a living
biological
system
Figure 10.15: The activities
within a processing
unit
Figure 10.16: Representation of
a processing unit
Figure 10.17: A neural network
with two different programs
(continued)
Figure 10.17: A neural network
with two different programs
Figure 10.18: Uppercase C and
uppercase T
Figure 10.19: Various
orientations of the letters
C and T (continued)
Figure 10.20: The structure of
the character
recognition
system
Figure 10.21: The letter C in the
field of view
Figure 10.22: The letter T in the
field of view
Figure 10.23: An artificial neural
network
implementing an
associative memory
Figure 10.24: The steps leading
to a stable
configuration
(continued)
Figure 10.24: The steps leading
to a stable
configuration
Figure 10.25: Crossing two
poker-playing
strategies
Figure 10.26: Coding the
topology of an artificial neural
network (continued)
Figure 10.26: Coding the
topology of an artificial neural
network
Figure 10.27: A semantic net
10.6 Applications of Artificial
Intelligence




Language processing.
Robotics.
Database systems.
Expert systems.