INTELLIGENT SYSTEMS
LECTURE 1 -
INTRO TO INTELLIGENT SYSTEMS
course introduction
GRACE (UMBRELLA)
grace is an intelligent agent who interacts with her environment
.
sensors -
cameras and microphones ; perceive and interact with the environment
·
actuators -
Speech and human-like actions ; perceive ,
reason about her options , represent knowledge ,
and adapt her behavior according to her expe-
rience .
intelligent methods :
·
perception
·
rationality
·
explicitly representing knowledge
·
learning and adapting
in this course we will not so much look into the question of perception via sensors ,
but mostly about the desicion making part ,
the rational behavior part
.
>
- ai and intelligent systems is not science fiction .
core topics
problem solving -
heuristics are one method of problem solving used by intelligent agents.
knowledge reasoning -
how we can represent information in a structured way . hierarchies of information are crucial for the
functioning of artificial intelligence .
·
adaptivity learning -
learning requires the combining of structured information gained from experience with problem solving methods .
.
e .
g using machine learning is used to make a horse look like a Zebra in motion . this is done by learning what is called latent features
of being a Zebra /namely having stripes) and propagate this on the video of a horse . You can train neural networks to be able to do this .
simpler applications of machine learning are the classification of emails are spam or regular mail .
schnapsen card
-points
ace 17
a popular card game in austria ,
southern germany ,
and hungary
.
ten 10
similar to 66
king 4
queen 3
it is a trick-based card game ,
Whoever wins starts the next trick.
jack 2
GOAL : collect 66 points >
- tallon or stock .
GAME EXPLANATIONS
·
there are only 20 cards used in schnapsen .
·
each player has five cards in their hand ,
and there is a stack of ten cards in the center .
·
in the first phase of the game ,
which last while the deck (tallon) in the Middle is not empty ,
where the player wins the trick if the play a
higher card.
·
in the second phase of the
game ,
the players are required to follow suit of the previous card When they play a card.
, PROPERTIES
adversarial agent setting
-
the opponent inter feres with your objective
·
imperfect information game
-
much uncertainty ,
e g
.
. What cards does the opponent have ?
2 game phases
·
explicit knowledge available
games are an adversarial setting , meaning I have an adversary which has a different goal than I do. that means the other
agent's actions interfere with your ability to reach your goal. It is an imperfect information game because we don't know the
order of the cards in the deck in the first phase of the game .
discrete mathematics
SETS
-
1 + 07
·
set notation -
[2 4 0 , 03 1 to 7
3
. ,
124,5
0 even integers 1 to 7
7 odd integers 3 to
·
elements -
2 4 ,
6 8 -
1
, , abs . Value = 1
W
·
subset -
AEB , 92 43 ,
is a subset of G2 ,
4 ,
6 , 03
·
disjoint -
[1 ,
2 ,
33 and 54 ,
5 , 63 are disjoint because they share no elements
empty set -
0 or 53
cardinality number of elements in IAI is Cardinality [1 is
·
-
the a set ; ,
where A the set .
the of , 2 , 33 3
.
·
power set -
all possible subsets of A , including the empty set and A itself .
If A has n elements ,
its powerset will have a "subsets.
·
Union -
AU B ,
set cointaining all the elements that are either in Set A ,
or set B ,
or in both . the union of [1 .
2 , 33 and 53 ,
4
, 53 is [1 .
2 ,
3 ,
4 ,
53 .
·
intersection -
A1 B ,
is the set containing all the elements that are both set A and Set B .
the intersection of 51 .
2 , 33 and 93 ,
4 . 53 is 933 .
·
disjointness -
A 1 B =
0 , disjoint sets are sets that have no elements in common .
N-TUPLES AND RELATIONS
n-tuples are sequences of length n ,
whose order matters Say ,
as , do , ..., an
(a) is a singlet ... 5- tuple
(a ,
b) is a pair ...
G-tuple
(a ,
b ,
c) is a triple ...
n-tuples
relations are sets of n-tuples Cordered) ,
C .
.
g
likes =
[(john , marx) , (john ,
cars) , ..., (Mary , apples)
in other words ,
subsets of the Cartesian product &john , mary3 x Emary ,
cars , applesy
GRAPS
undirected graph
·
a non-empty set of vertices V ,
and a set of unordered ,
distinct edges E
.
① GV ES v =G 4 r my
/
. . ,
E =
GEU vy Er my En was
. . , , .
,directed
·
a non-empty set of vertices V ,
and a set of directed edges E , consisting of ordered pairs.
&
VE van
E =
G(n v)
. .
(v m)
, , (n m)]
.
TREES
a parent node is any node which has children ,
which are nodes that appear below it in a tree and can be directly expanded towards , or reached ,
from the
parent . this node is only called the parent when there is a direct link to the child using one edge, not an extended path .
an internal node is one which has both a parent and children) .
in contrast ,
leaf nodes have a parent but no child(ren) .
Nodes are
siblings when they share the same parent node .
ancestors to a node are all nodes which must be passed through when drawing the shortest path to the root .
In this example ,
the path
from hand i to the root node a must pass through the nodes e ,
b ,
and a
ROOT NODE
INTER
&
PARENT OF
6 AND F
n
C
-
⑳ + 9
XX
n i
0 ANCESTORS OF HAND I
RATIONAL AGENTS
eceptaso
the
agent function maps percent sequence to actions
r~
&
*
f : p - A
environment
the agent function will internally be represented by the agent program.
actions e actuators
the agent program runs on the physical architecture to produce f.
some entity (the "agent") which consists of a set of sensors and actuators . With the sensors the agent perceives the environment ,
and
With actuators it can manipulate the environment through a set of actions .
the levels of abstraction one can study agents ,
1. functional-agent behaviour as a function with sensor information as input le .
g . as a rector of features) and a
possibly complex) action as input
.
2
. the implementation in some
program system
-
.
3 the Specific architecture in which the agent operates .
RATIONALITY
a rational agent chooses which ever action maximizes the expected value of the performance measure given the percept sequence to date
, and prior environment knowledge .
what is rational at a given time depends :
·
expected value of performance measure /heuristic)
·
actions and choices (search
prior and environment knowledge (knowledge representation
·
percept sequence to date (learning)
rationality is not the same as omniscience or perfection .
agent types
SIMPLE REFLEX
select action on the basis of only the current percept pseudocode for the agent function :
implement through condition-action rules . function REFLEX-AGENT-WITH-STATE (percept) returns an action
*
Static : rules ,
a set of condition -
action rules
state ,
a description of the current world state
action ,
the most recent action .
State - UPDATE-STATE /state ,
action ,
percept)
rule -
RULE-MATCH /State ,
rule)
action -
RULE-ACTION [rule]
return action
MODEL BASED
to tackle partially observable environments .
·
maintain internal state .
over time update state using world knowledge
does world ?
how the change
·
·
how do actions affect world ? >
- model of world
·
GOAL BASED
the agent needs a goal to know which situations are desirable.
typically investigated in search and planning research .
major difference :
the future is taken into account .
·
LECTURE 1 -
INTRO TO INTELLIGENT SYSTEMS
course introduction
GRACE (UMBRELLA)
grace is an intelligent agent who interacts with her environment
.
sensors -
cameras and microphones ; perceive and interact with the environment
·
actuators -
Speech and human-like actions ; perceive ,
reason about her options , represent knowledge ,
and adapt her behavior according to her expe-
rience .
intelligent methods :
·
perception
·
rationality
·
explicitly representing knowledge
·
learning and adapting
in this course we will not so much look into the question of perception via sensors ,
but mostly about the desicion making part ,
the rational behavior part
.
>
- ai and intelligent systems is not science fiction .
core topics
problem solving -
heuristics are one method of problem solving used by intelligent agents.
knowledge reasoning -
how we can represent information in a structured way . hierarchies of information are crucial for the
functioning of artificial intelligence .
·
adaptivity learning -
learning requires the combining of structured information gained from experience with problem solving methods .
.
e .
g using machine learning is used to make a horse look like a Zebra in motion . this is done by learning what is called latent features
of being a Zebra /namely having stripes) and propagate this on the video of a horse . You can train neural networks to be able to do this .
simpler applications of machine learning are the classification of emails are spam or regular mail .
schnapsen card
-points
ace 17
a popular card game in austria ,
southern germany ,
and hungary
.
ten 10
similar to 66
king 4
queen 3
it is a trick-based card game ,
Whoever wins starts the next trick.
jack 2
GOAL : collect 66 points >
- tallon or stock .
GAME EXPLANATIONS
·
there are only 20 cards used in schnapsen .
·
each player has five cards in their hand ,
and there is a stack of ten cards in the center .
·
in the first phase of the game ,
which last while the deck (tallon) in the Middle is not empty ,
where the player wins the trick if the play a
higher card.
·
in the second phase of the
game ,
the players are required to follow suit of the previous card When they play a card.
, PROPERTIES
adversarial agent setting
-
the opponent inter feres with your objective
·
imperfect information game
-
much uncertainty ,
e g
.
. What cards does the opponent have ?
2 game phases
·
explicit knowledge available
games are an adversarial setting , meaning I have an adversary which has a different goal than I do. that means the other
agent's actions interfere with your ability to reach your goal. It is an imperfect information game because we don't know the
order of the cards in the deck in the first phase of the game .
discrete mathematics
SETS
-
1 + 07
·
set notation -
[2 4 0 , 03 1 to 7
3
. ,
124,5
0 even integers 1 to 7
7 odd integers 3 to
·
elements -
2 4 ,
6 8 -
1
, , abs . Value = 1
W
·
subset -
AEB , 92 43 ,
is a subset of G2 ,
4 ,
6 , 03
·
disjoint -
[1 ,
2 ,
33 and 54 ,
5 , 63 are disjoint because they share no elements
empty set -
0 or 53
cardinality number of elements in IAI is Cardinality [1 is
·
-
the a set ; ,
where A the set .
the of , 2 , 33 3
.
·
power set -
all possible subsets of A , including the empty set and A itself .
If A has n elements ,
its powerset will have a "subsets.
·
Union -
AU B ,
set cointaining all the elements that are either in Set A ,
or set B ,
or in both . the union of [1 .
2 , 33 and 53 ,
4
, 53 is [1 .
2 ,
3 ,
4 ,
53 .
·
intersection -
A1 B ,
is the set containing all the elements that are both set A and Set B .
the intersection of 51 .
2 , 33 and 93 ,
4 . 53 is 933 .
·
disjointness -
A 1 B =
0 , disjoint sets are sets that have no elements in common .
N-TUPLES AND RELATIONS
n-tuples are sequences of length n ,
whose order matters Say ,
as , do , ..., an
(a) is a singlet ... 5- tuple
(a ,
b) is a pair ...
G-tuple
(a ,
b ,
c) is a triple ...
n-tuples
relations are sets of n-tuples Cordered) ,
C .
.
g
likes =
[(john , marx) , (john ,
cars) , ..., (Mary , apples)
in other words ,
subsets of the Cartesian product &john , mary3 x Emary ,
cars , applesy
GRAPS
undirected graph
·
a non-empty set of vertices V ,
and a set of unordered ,
distinct edges E
.
① GV ES v =G 4 r my
/
. . ,
E =
GEU vy Er my En was
. . , , .
,directed
·
a non-empty set of vertices V ,
and a set of directed edges E , consisting of ordered pairs.
&
VE van
E =
G(n v)
. .
(v m)
, , (n m)]
.
TREES
a parent node is any node which has children ,
which are nodes that appear below it in a tree and can be directly expanded towards , or reached ,
from the
parent . this node is only called the parent when there is a direct link to the child using one edge, not an extended path .
an internal node is one which has both a parent and children) .
in contrast ,
leaf nodes have a parent but no child(ren) .
Nodes are
siblings when they share the same parent node .
ancestors to a node are all nodes which must be passed through when drawing the shortest path to the root .
In this example ,
the path
from hand i to the root node a must pass through the nodes e ,
b ,
and a
ROOT NODE
INTER
&
PARENT OF
6 AND F
n
C
-
⑳ + 9
XX
n i
0 ANCESTORS OF HAND I
RATIONAL AGENTS
eceptaso
the
agent function maps percent sequence to actions
r~
&
*
f : p - A
environment
the agent function will internally be represented by the agent program.
actions e actuators
the agent program runs on the physical architecture to produce f.
some entity (the "agent") which consists of a set of sensors and actuators . With the sensors the agent perceives the environment ,
and
With actuators it can manipulate the environment through a set of actions .
the levels of abstraction one can study agents ,
1. functional-agent behaviour as a function with sensor information as input le .
g . as a rector of features) and a
possibly complex) action as input
.
2
. the implementation in some
program system
-
.
3 the Specific architecture in which the agent operates .
RATIONALITY
a rational agent chooses which ever action maximizes the expected value of the performance measure given the percept sequence to date
, and prior environment knowledge .
what is rational at a given time depends :
·
expected value of performance measure /heuristic)
·
actions and choices (search
prior and environment knowledge (knowledge representation
·
percept sequence to date (learning)
rationality is not the same as omniscience or perfection .
agent types
SIMPLE REFLEX
select action on the basis of only the current percept pseudocode for the agent function :
implement through condition-action rules . function REFLEX-AGENT-WITH-STATE (percept) returns an action
*
Static : rules ,
a set of condition -
action rules
state ,
a description of the current world state
action ,
the most recent action .
State - UPDATE-STATE /state ,
action ,
percept)
rule -
RULE-MATCH /State ,
rule)
action -
RULE-ACTION [rule]
return action
MODEL BASED
to tackle partially observable environments .
·
maintain internal state .
over time update state using world knowledge
does world ?
how the change
·
·
how do actions affect world ? >
- model of world
·
GOAL BASED
the agent needs a goal to know which situations are desirable.
typically investigated in search and planning research .
major difference :
the future is taken into account .
·
