Artificial Intelligence

1
Artificial Intelligence
2
Artificial Intelligence
Artificial intelligence (AI) is a broad field,
and means different things to different people.
The field of Artificial Intelligence (AI) is
concerned both with modeling human intelligence
and with solving complex problems not solvable by
simple or analytic procedures. For instance, a
major goal of AI is construction of an
intelligent robot, capable of perceiving, acting,
comprehending, reasoning, and learning in complex
environments.
alison_at_ Fri Aug 19 104217 BST 1994
3

• The AI field consists of six related areas
• Problem solving search
• Knowledge Representation
• Natural Language Processing (NLP)
• Reasoning Systems
• Vision Perception

4

• Problem Solving Search
• A fundamental technique in AI is to encode a
  problem as a state space in which solutions are
  goal states in that space.
• Thus, problem solving can be viewed as state
  space search.
• To search large, combinatorial state spaces,
  knowledge (e.g. heuristics) and planning are
  required.

5
Search Often there is no direct way to find a
solution to some problem. However, you do know
how to generate possibilities. For example, in
solving a puzzle you might know all the possible
moves, but not the sequence that would lead to a
solution. When working out how to get somewhere
you might know all the roads/buses/trains, just
not the best route to get you to your destination
quickly. Developing good ways to search through
these possibilities for a good solution is
therefore vital. Brute force techniques, where
you generate and try out every possible solution
may work, but are often very inefficient, as
there are just too many possibilities to try.
Heuristic techniques are often better, where you
only try the options which you think (based on
your current best guess) are most likely to lead
to a good solution.
6
Searches Breadth First Search Depth First
Search Heuristic Search Generate and Test
Technique Simple Hill Climbing Steepest-Ascent
Hill Climbing Best First Search A Search
Genetic Searches
7

• Knowledge Representation
• Intelligent behavior often requires knowledge.
• For example, language comprehension requires
  encoding the meanings of words and how they are
  combined.
• Techniques for representing knowledge
  include use of semantic networks, logic, and
  neural networks.

8
Multi-valued Logic
Language
First Order Logic
Propositional Logic
Pseudo-Boolean
Agents
KNOWLEDGE REPRESENTATION
Modeling
Hierarchies
Incomplete Knowledge
Model Clustering
Supermodels
Solution Structure
Symmetry
9
Semantic Nets The simplest kind of structured KE
is the semantic net originally developed in the
early 1960s to represent the meaning of English
words. A semantic net is really just a graph,
where the nodes in the graph represent concepts,
and the arcs represent binary relationships
between concepts. Predicate Logic The most
important knowledge representation language is
arguably predicate logic (or strictly, first
order predicate logic - there are lots of other
logics out there to distinguish between). It is
a well-defined syntax, semantics and rules of
inference. . Other logics
10

• Natural Language Processing
• Language is the major medium for communicating
  thought and knowledge.
• NLP is concerned with mappings between language
  and thought, how language skills are learned, and
  how knowledge is acquired through language (e.g.
  reading).
• Really understanding a single sentence requires
  extensive knowledge both of language and of the
  context.
• For example They can fish can only be
  interpreted reasonably if you know the context,
  some ones job (canning fish) or recreational
  activity.

11

• Reasoning Systems
• Most human reasoning occurs in task/domains with
  uncertain, ill-defined and incomplete knowledge.
• Reasoning in such domains requires techniques
  such as use of default, probabilistic, and
  non-monotonic logics.
• Expert Systems
• Fuzzy Logic
• Case-Based Reasoning
• Neural Networks

12

• Expert Systems
• most common application utilized today employing
  the knowledge of an "expert" to solve problems or
  make decisions.
• Components of Expert Systems
• Expert in the Field - to establish knowledge base
  on topic
• Knowledge Engineer - Interviews experts and
  establishes (Heuristic) rules in an IF-THEN
  manner
• Knowledge Base - Information stored on the
  computer about subject matter
• Inference Engine - Compares inputs to known set
  of rules
• Benefits
• Provides information quicker than its human
  counterparts
• Utilizes standards to diagnose relatively
  consistent problems
• Limitations
• Systems are not flexible changing the knowledge
  base can be cumbersome
• Not applicable for problems that don't follow the
  rules

13
Rule-Based Systems Consists of IF-THEN rules,
facts, and some interpreter (chainer) Two basic
types forward chaining and backward chaining
systems. Forward chaining systems start with
the initial facts, and keep using the rules to
draw new conclusions (or take certain actions)
given those facts. DATA DRIVEN Backward
chaining systems start with some hypothesis (or
goal) you are trying to prove, and keep looking
for rules that would allow you to conclude that
hypothesis to be true, perhaps setting new
subgoals to prove as you go. GOAL DRIVEN
14
Uncertainty in Rules Rules look pretty much like
logical implications. In practice you rarely
conclude things with absolute certainty. Usually
we want to say things like If Alison is tired
then there's quite a good chance that she'll be
in a bad mood''. To allow for this sort of
reasoning in rule-based systems we often add
certainty values to a rule, and attach
certainties to any new conclusions. We might
conclude that Alison is probably in a bad mood
(maybe with certainty 0.6). The approaches used
are generally loosely based on probability
theory, but are much less rigorous, aiming just
for a good guess rather than precise
probabilities.You can tune your expert system
with these certainty factors.
15

• Fuzzy Logic
• Problems we encounter in our daily work and home
  environments frequently lack a complete black or
  white answer. Instead theres a level of grayness
  - a degree of rightness or wrong-ness. It is for
  this that Fuzzy Logic was developed.
• Components of Fuzzy Logic Systems
• Arithmetic Logic Unit - to interpret
  relationships such as ,lt,, etc.
• Set of Knowns - To compare an input to Membership
  Function
• Calculation that determines how closely the input
  can be matched to a set of knowns
• Benefits
• Provides information quicker than its human
  counterparts
• Increased flexibility over Expert Systems Less
  rigid
• Limitations
• Very program intensive to make robust
• More costly than expert system alternative

16

• Case Based Reasoning
• Case-Based Reasoning program intensive and
  costly
• As with expert systems and fuzzy logic, there are
  instances when the information regarding a
  decision in question does not fit into any of the
  structured rules so in these situations, the
  facts and past decisions with similar
  characteristics become important
• Components of Case-Based Reasoning Systems
• Knowledge Base (casebase) Information about
  subject matter
• Inference Engine - Compares inputs to known cases
  with solutions
• Random Access Type Memory - Allows new cases to
  be indexed thus increasing knowledge (case) base
• Benefits
• Ability to obtain solutions in areas not
  previously well understood
• Limitations
• Little quality control or guarantee that the
  solution is optimum.
• Costly development.

17

• Neural Network
• Ultimate goal of AI is to imitate human thought--
  artificial neural networks attempt to replicate
  the connectivity and functioning of biological
  neural networks (i.e. the human brain). Theory
  is that replicating the brains structure, the
  artificial network will, in turn, possess the
  ability to learn.
• Components of Neural Networks
• Network of Nodes (either physical or virtual) -
  interconnected with one another that sense and
  process data
• Layered Structure - that increases sensitivity
  and accuracy of the transmission between nodes
  and carry out specific functions
• Benefits
• The Ability to learn , Flexibility to handle any
  type of problem
• Limitations
• Very sensitive and must be fully trained

18
Neural Network (Learning) System General pattern
classification pattern recognition pattern
function estimation/approximation Agent Planning
System (household or work robot) specification
of environment, tasks, and actions model of
environment KB for environment and actions
planning system goal setting / evaluation
system examples monitoring security household
robot autonomous housekeeping robot command
controlled household servant (with
NLI?)combination of above (same for
factory/plant environment)
19
Real neurons Figure of biological neuron
                                                  
                                              
20
Artificial neurons Figure of TLU
21

• Vision and Perception
• Images are fraught with ambiguity, e.g., wiggly
  lines could represent ocean waves, a person's
  hair, snakes, etc.
• Low-level vision is concerned with extracting
  visual features from color, texture, edges, and
  so forth, while high-level vision deals with how
  to represent and form internal models of complex
  shapes and structured objects.

22
Image processing is in many cases concerned with
taking one array of pixels as input and producing
another array of pixels as output which in some
way represents an improvement to the original
array. Image processing methods may be broadly
divided into Real space methods -- which work
by directly processing the input pixel array.
Fourier space methods -- which work by firstly
deriving a new representation of the input data
by performing a Fourier transform, which is then
processed, and finally, an inverse Fourier
transform is performed on the resulting data to
give the final output image.
23
(No Transcript)