Introduction and
problem solving
11/01/2014
Unit 1
Lesson 1
Introduction to
Artificial Intelligence
3 [19659002] Contents
1. What is artificial intelligence?
2. Historical perspective
3. Types of systems IA
4. Framework I + D + I (National and European)
4
Objectives
Get a broad vision of the concept
Artificial Intelligence
Know the historical framework of Intelligence
Artificial [19659002] Taxonomy and types of systems within the
Artificial Intelligence
What is Artificial Intelligence?
5
"The study of the mental faculties through the use of
computational models" ( Charniaky McDermott, 1985)
"IA (…) is related to intelligent behaviors in artifacts"
(Winston, 1992).
"The study of how to get computers to perform tasks
that, for the moment, humans do better "(Rich and Knight,
1991).
" Capable of understanding, assimilating, elaborating information and using it
properly. "
What is Artificial Intelligence? [19659002] 6
Linguistics
IA
Psychology
Informatica
7
What is Artificial Intelligence?
• Types of systems in AI:
• Systems based on knowledge
• Definition of such knowledge through an expert [19659002] • Systems based on learning
• Knowledge can be learned from concrete cases and
examples
• Are treated differently and the techniques used
also differ but can complement the one
other
8
What is Artificial Intelligence?
• Dimensions of AI research:
• Development of new functionalities: focuses on solving
problems through the use of computing that until now or
otherwise can not be resolved (eg OCR, recognition
of objects, etc.).
• Methods and tools used in systems: environments of
development of expert systems (eg CLISP, Prolog).
• Development and use in applications commercial: cycles of 5-5-5
(research-development-dissemination).
9
Historical perspective
• Foundations (400 BC)
• Aristotle (384-322 BC) Understanding through reason
• Formalization of al-Khowarazmi algorithms (9th century)
• first calculation machines by Pascal
(17th century) and Charles
Babagge (s. XIX)
• First computers (about 1940)
• Z-3 was invented by Konrad Zuse in 1941. In the United Kingdom, the
first system by Alan Turing in 1940, and the Colossus in 1943. In
United States, the ABC between 1940 and 1942 by John Atanasoff
• ENIAC 1946 (Electronic Numerical
Integrator
and
Computer.)
10
Historical Perspective
• Genesis (1943-1956)
• McCulloch and Pitts have been recognized as the authors of the
first IA work , in 1943, proposing a model constituted
by artificial neurons
• First chess game programs developed by
Shannon and Turing between 1950
• first machine translation systems, such as experiment
Georgetown-IBM
Darmouth Seminar 1956 origin of the
term artificial intelligence
Bake Library, University of Darmouth.
11
Historical perspective
• Initial enthusiasm, great hopes (1952-1969) )
• Great success at both university and business level and
research centers
• Development of tools such as LISP
• Crisis / redimensioning of problems (1966-1973)
• Li mitations computations and hardware are not the only
problem
• Decrease in funding and a large number of
projects are discontinued
Historical perspective
12
• Resurgence
in
centered
knowledge (1969-1979)
• A renewed interest arises for expert systems based on the
knowledge applied to domains such as medical diagnosis or control
of plants
systems [19659002] based
on
on
• The AI industry (1980 to the present)
• Expert systems begin to report benefits in their various
applications (eg DEC order systems)
• New aspects of AI begin to develop such as mining
data or semantic technology
• 2013- ….
• Real-time and ubiquitous analysis and learning systems for
adapt to the variability and needs of the environment (Big Data).
13
Taxonomy of the AI
• Knowledge-based systems
• Has a knowledge base
• Need of acquisition, formalization, and coding
• Expert systems belong to this category
system
expert
is
knowledge
computerized
a
A
that
system
a
uses
a
domain
that
problem
domain,
the
solution must be essentially
the same as the one provided
by a human expert in that
domain. [19659002] of
solve
of
than
for
specific
form
of
14
Taxonomy of AI
• Systems based on learning
• Consists of a training phase and an operational phase
• The training phase used examples to create a model
• The operational phase executes the learned model to take
decisions. [19659002] • OCR or image object recognition systems are two
examples of this type of systems
15
Taxonomy of AI
• Neural Networks, Bayesian Networks, Genetic Algorithms and [19659002] Case-based Reasoning
Experimental applications and systems
experimental
16
• Railway transport control systems
• Diagnostic and repair systems (medicine and
automotive)
] • Agents
intelligent
in [19659002] video games
and
trade
electronic
• Data mining is common in astronomy, biology,
remote sensing, web content analysis
• Fraud detection through pattern analysis
Experimental applications and systems
experimental
17
• Prior to commercialization a cycle of 5-5-5
(research-development-commercialization)
is produced. references are:
• MIT (Massachusetts Institute of Technology)
• Standford KSL
• Xerox PARC
• ATT Labs
• IBM Watson Laboratory
• Research Institute in Artificial Intelligence of the Superior Council
of Scientific Investigations
• Many universities also have an AI group (in
the European University was created in 1999 the group of
research in Intelligent Systems te)
18
Abstract
• Focused on this subject as Computer Science
Advanced Applied and Multidisciplinary Discipline
• Historical Perspective
• Fundamentals (400 ac!).
• First computers (about 1940).
• Genesis (1943-1956).
• Initial enthusiasm, great hopes (1952-1969).
• Crisis / redimensioning of problems (1966-1973). • The AI industry (1980 to the present).
• Darmouth Seminar in 1956 as the origin of the
term
• Systems based on knowledge and systems
based on learning
Unit 1
Lesson 2
Resolution of
problems using
search
20
Contents
1. Problem solving by
abstraction
2. Formulation of problems as spaces of
states
3. Search trees and search for
solutions
21
Objectives
Define a problem from the computational
point of view
Find the actions that lead to the solution
22
Problems and abstraction
• Abstraction is a fundamental part in various
techniques and methods of computing in general
By abstract we usually understand focusing only on the
aspects of a problem that we identify as main,
knowing how to leave aside a lot of details that we decided not
are relevant.
• Example case: we want to go from one city to another in
Romania (from Arad to Bucharest)
] Problems and abstraction
23
24
Formulation of the problem as a
space of states
• There are different forms of formalization, as can
be a quintuple: P = [E, Ei, O,M, C] .
Formulation of the problem as a
state space
25
26
Example: robot vacuum cleaner (I)
• Build a control system of a vacuum cleaner, which has
an engine to vacuum and another to move from one
place to another.
one
of
How many possible states
in
two
units where each
can be dirty or clean?
space
2 robot positions * 2 possible
states for cell1 * 2
possible states for cell2
= 8
27
Example: robot vacuum cleaner (II)
P = [E, Ei, O,M, C]
E =
Ei = {E1}
O = {OI, OD, OS}
M = {E7, E8}
C = {1 , 1, 1}
Example: robot vacuum cleaner (III)
28
What is the minimum cost?
What is the maximum cost?
Example II: the puzzle (I)
29
Formalization in quintuplets: [E, Ei, O,M, C]
How many possible states?
181,440 possible states
30
Example II: the puzzle (II)
• E = {E1, ….. En}: n = 9! / 2 (half of the states are mirror)
• Ei = {E1}
Different ways of defining an operator
• O = {OA (1), OA (2), … OD (1) .. ..OI (1), …. OAb (1) … OAb (8)}
• O = {OA (h), OD (h), OI (h), Oab (h)}
Move the pieces
Move the hole
• M = {E2}
• C = {1, 1, … 1}
31
Search and search trees
solutions (I)
• To solve the problem it is necessary to have a
algorithm that allows to travel through the different states
to get from the initial state to the desired final state
• This algorithm is based on a tree structure for
perform search and so called tree of
search
• Operators are executed in the initial state and
get the events beef. Subsequently apply another
times the operators with these and so on until
find the goal
Search and search trees of
solutions (II)
32
33
Trees search and search for
solutions (III)
• The traveler wanted to go from Arad to Bucharest
• One solution is [AradZerindOradeaSibiuFagaras
Bucharest]
Search and search trees
solutions (IV)
34
35
search trees and search for
solutions (V)
search-tree function returns a solution or failure
loop do
if not there are candidates to expand then return failure
choose, according to strategy, a leaf node to expand
if the node contains a target state then return the
corresponding solution
otherwise expand the node and add the result nodes to
search tree
final loop
Final function
What is the tree's expansion order? What type of tree is generated?
36
Summary
• Techniques for modeling problems through the
from Nettech Post http://bit.ly/2s0XCgM
No hay comentarios:
Publicar un comentario