1 rbs cs 331/531 dr m m awais production systems symbolic based rule based system based on simple...

CS 331/531 Dr M M Awais 1

RBS

Production Systems

Symbolic based Rule based systemBased on simple working principleSystem consists of:

production rulesworking memoryrecognise-act control cycle.


RBS

Production Rules (Productions)

Production Rules are:

Condition – Action Pair

Defines the associated problem-solving step.

Pattern that determines when to apply rule


RBS

Working Memory

Contains the description of current state

WM state is compared with the condition part of the production rules to determine the type of action required.

Always changes when productions are applied


RBS

Recognise – Act CycleRecognise – Act Cycle

Initialise Working Memory

Productions - Conditions

Conflict Set

Conflict Resolution

• (if the two match)

(one pattern taken at a time)

(Enabled)

Production - (Fired)

Match the WM memory with the conditions of rulesMatch the WM memory with the conditions of rules


RBS




Conflict Set

Conflict Resolution



(Enabled)


Productions can only fire when conditions in production-conditions

are matched in the working memory (WM).

If they match a subset of productions is formed called Conflict Set.

Part of Production That Are EnabledPart of Production That Are Enabled


RBS




Conflict Set

Conflict Resolution



(Enabled)


It chooses a rule from conflict set for firing. Selection of any rule It chooses a rule from conflict set for firing. Selection of any rule may be based on complex rule selection heuristics.may be based on complex rule selection heuristics.


RBS




Conflict Set

Conflict Resolution



(Enabled)


Fire a rule and update the WMFire a rule and update the WM


RBS

Feedback in the Systems in which no conflict resolution Feedback in the Systems in which no conflict resolution remains. All conflicts are resolvedremains. All conflicts are resolved

A Production System

WMWMPatternsPatterns

P A1P A2

----

Pattern Action


RBS

Example:Production Set:

1.1. ba ba ab ab

2.2. ca ca ac ac

3.3. cb cb bc bc(whenever a pattern on the LHS appears replace it with (whenever a pattern on the LHS appears replace it with

the one on the RHS )the one on the RHS )

Working Memory = acbccWorking Memory = acbcc

Conflict Set = 3

Rule fired = 3

acbcbcc abcbccc


RBS Example:Production Set:

1.1. ba ba ab ab

2.2. ca ca ac ac

3.3. cb cb bc bc


RBS Example:Production Set: 1.1. ba ba ab ab

2.2. ca ca ac ac

3.3. cb cb bc bc

Iteration#

WM Conflict Set

Rule Fired

0 cbaca 1,2,3 1

1 cabca 2 2

2 acbca 2,3 2

3 acbac 1,3 1

4 acabc 2 2

5 aacbc 3 3

6 aabcc NIL NIL



2.2. ca ca ac ac

3.3. cb cb bc bc

Iteration#

WM Conflict Set

Rule Fired

0 cbaca 1,2,3 1

1 cabca 2 2

2 acbca 2,3 2

3 acbac 1,3 1

4 acabc 2 2

5 aacbc 3 3

6 aabcc NIL NIL


RBS

Production Set:

1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. Start v ^ r ^ qIF symbols on the LHS appear then add the

symbols on the RHS to the WM

Data Driven Search in Production System


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. start v ^ r ^ q

Data Driven: Production Set:

Iteration#

WM Conflict Set

Rule Fired

0 start 6 6

1 start,v,y,q 6,5 5

2 start,v,r,q,s 6,5,2 2

3 start,v,r,q,s,p

6,5,2,1 1

4 start,v,r,q,s,p, goal

6,5,2,1 NIL


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. start v ^ r ^ q


Iteration#

WM Conflict Set

Rule Fired

0 start 6 6

1 start,v,r,q 6,5 5

2 start,v,r,q,s 6,5,2 2

3 start,v,r,q,s,p

6,5,2,1 1


6,5,2,1 NIL


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. start v ^ r ^ q


Iteration#

WM Conflict Set

Rule Fired

0 start 6 6

1 start,v,y,q 6,5 5

2 start,v,r,q,s 6,5,2 2

3 start,v,r,q,s,p

6,5,2,1 1


6,5,2,1 NIL


RBS

start

qrv


RBS

start

qrv

s


RBS

start

qrv

s

p


RBS

start

qrv

s

p

goal


RBS

Production Set:

1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. Start v ^ r ^ q

GOAL Driven Search in Production System


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. start v ^ r ^ q

Goal Driven: Production Set:

Iteration# WM Conflict Set Rule Fired

0 Goal 1 1

1 Goal,p,q 1,2,3,4 2

2 Goal,p,q,r,s 1,2,3,4,5 3

3 Goal,p,q,r,s,w 1,2,3,4,5 4

4 Goal,p,q,r,s,w,t,u 1,2,3,4,5 5

5 Goal,p,q,r,s,w,t,u,v 1,2,3,4,5,6 6

6 Goal,p,q,r,s,w,t,u,v,start

1,2,3,4,5,6 NIL


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r q

4. t ^ u q

5. v s

6. start v ^ r ^ q



0 Goal 1 1

1 Goal,p,q 1,2,3,4 2

2 Goal,p,q,r,s 1,2,3,4,5 3

3 Goal,p,q,r,s,w 1,2,3,4,5 4

4 Goal,p,q,r,s,w,t,u 1,2,3,4,5 5

5 Goal,p,q,r,s,w,t,u,v 1,2,3,4,5,6 6


1,2,3,4,5,6 NIL


RBS 1. p ^ q goal

2. r ^ s p

3. w ^ r p

4. t ^ u q

5. v s

6. start v ^ r ^ q



0 Goal 1 1

1 Goal,p,q 1,2,3,4 2

2 Goal,p,q,r,s 1,2,3,4,5 3

3 Goal,p,q,r,s,w 1,2,3,4,5 4

4 Goal,p,q,r,s,w,t,u 1,2,3,4,5 5

5 Goal,p,q,r,s,w,t,u,v 1,2,3,4,5,6 6


1,2,3,4,5,6 NIL


RBS

goal

qp


RBS

goal

qp

r

w


RBS

goal

qp

r

w

sr


RBS

goal

qp

r

w

sr

t

u


RBS

goal

qp

r

w

sr

t

u

v


RBS

goal

qp

r

w

sr

t

u

start

v


RBS

Hearsay – II (Erman et.al 1980)

• One of the first developed project

• AIMED at speech recognition system


RBS

Rule Based System Modeled as Production systems Production Systems implemented as

Cause-Effect rules RBS have IF-THEN rules in their

structures Working principle is same as

production system Capable of handling vague concepts


RBS

Rule Based System Find all the rules which have the IF

part satisfied Select one rule using conflict

resolution Perform action and modify the

Working memory Repeat until all the rules have been

fired or the goal have been reached.


RBS

Example: RBS with 5 Rules R1: If hot and smoky then add fire R2: If alarm beeps then add smoky R3: If fire then add switch on the sprinklers R4: If dry Then add switch on humidifier R5: If switch on sprinklers then Delete dry

Given Facts: Alarm beeps, hot, dry

If you associate every fact with symbols it becomes a production system


RBS

Forward Chaining:Rule firing

R2 and R4 can be fired If R2 fired then

R1, R3, and R5 will follow As a result: Dry will be deleted from WM

If R4 is fired R2, R1, R3, and R5 will be fired As a result: Dry will be deleted from WM

R4 is fired AimlesslyR4 is fired Aimlessly

R1: If hot and smoky then add fire

R2: If alarm beeps then add smoky

R3: If fire then add switch on the sprinklers

R4: If dry Then add switch on humidifier

R5: If switch on sprinklers then Delete dry



RBS


Alarm beeps Hot dry








RBS


Alarm beeps Hot dry

Switch_on_humdifiers

R4

smoky

R2








RBS


Alarm beeps Hot dry


R4

smoky

R2

fire

R1








RBS


Alarm beeps Hot dry


R4

smoky

R2

fire

R1

Switch_on_sprinklers

R3








RBS


Alarm beeps Hot dry


R4

smoky

R2

fire

R1


R3







Delete dry


RBS


Alarm beeps Hot dry


R4

smoky

R2

fire

R1


R3







Delete dry

Invalid rule


RBS

Forward Chaining

End up firing more than required rules

Generally applied when lots of facts are known about the system


RBS

Backward chaining From the goal find the initial

conditions that satisfy the goal.

Goals are divided into sub-goals and then further sub-goals are found until the initial conditions are reached

Applied when the goal is exactly known


RBS

Backward chaining Hypothesize that the goal is

true and then try to prove the hypothesis

Start from the GOAL and see if the WM supports the facts required to make the goal true


RBS

Example R1: if smoky and hot then add fire R2: If alarm beeps then add smoky R3: If alarm beeps then add ear plugs R4: If fire then add switch on sprinklers R5: If smoky then add poor visibility

Facts: Alarm beeps, hot Goal: Switch on sprinklers


RBS

Example

R1: if smoky and hot then add fire R2: If alarm beeps then add smoky R3: If alarm beeps then add ear plugs R4: If fire then add switch on

sprinklers R5: If smoky then add poor visibility


Switch on sprinklers

fire

R4

smoky hot

R1

Alarm beeps

R2No rule available but present in the facts


RBS

Example





fire

R4

smoky hot

R1

Alarm beeps



RBS

Example





fire

R4

smoky hot

R1

Alarm beeps


No rule available but present in the facts


RBS

Example





fire

R4

smoky hot

R1

Alarm beeps


No rule available but present in the facts

HENCE THE GOAL IS SATISFIEDHENCE THE GOAL IS SATISFIED


RBS Example IF (lecturing X) and (marking practicals X)

Then Add (overworked X) If (month feb) then add (lecturing ali) If (month feb) Then add (marking practical

ali) IF (overworked X) or (sleepless X) Then

(bad mood X) If (bad mood X) then Delete (happy X) If (lecturing X) the Delete (researching X) IF (year 2004) then (economy bad)

Fact Set 1: (month feb), (happy ali), (researching ali)Fact set 2: (month feb), (year 2004)Goal: find whether Ali is in bad mood or not


RBS

Conflict Resolution General Approaches:

General Conflict Resolution strategies Problem Specific Conflict Resolution

Strategies

Rules that control the resolution process are called META rules


RBS

General Conflict Resolution strategies: META Rules

Law of Recency Prefer rules for which the facts have

been added recently Law of Specificity

Prefer more specific rules, i.e, rules that have more conditions


RBS

Examples: Recency Rule1: Credit History Rule:IF X has had credit for less than 3 yearsTHENX’s credit history is very low Rule2: High Risk RuleIF X’s credit history is very lowANDX has been unemployed for more than half of his adult

yearsTHENX is a high credit risk customer Rule 3: Age RuleIF X’s age is less than 25 yearsTHENX is young


RBS Examples: Recency

Rule1: Credit History Rule:IF X has had credit for less than 3 yearsTHENX’s credit history is very low Rule2: High Risk RuleIF X’s credit history is very lowANDX has been unemployed for more than half of his adult


1. Atif is 25 years old2. Atif has 1 yr of credit history3. Atif unemployed for 3/5

of his adult life






of his adult life

1. Atif is 25 years old2. Atif has 1 yr of credit history3. Atif unemployed for 3/5 of hisadult life4. Atif’s credit history is very low

R1

Newly added fact

Why wasn’t the Rule 3 fired?, will be answered later






of his adult life

1. Atif is 25 years old2. Atif has 1 yr of credit history3. Atif unemployed for 3/5 of hisadult life4. Atif’s credit history is very low

R1

Newly added factR2

To be added: Atif is a high credit risk customer


RBS

Specificity Prefer rules with more specific

conditions I.e., rules having more antecedents R1: if smoky and hot then add fire

R2: If alarm beeps and hot then add smoky

R3: If alarm beeps then add ear plugs

R4: If fire then add switch on sprinklers

R5: If smoky then add poor visibility


Which rule is preferred R2 or R3?

R2 is more specific, so prefer R2


RBSProblem Specific Conflict Resolution strategies: META Rules

Prefer rules with credit history over others

Rule1: Credit History Rule:IF X has had credit for less than 3 yearsTHENX’s credit history is very low Rule2: High Risk RuleIF X’s credit history is very lowANDX has been unemployed for more than half of

his adult yearsTHENX is a high credit risk customer Rule 3: Age RuleIF X’s age is less than 25 yearsTHENX is young


RBS

Decision Trees Vs RBS

Decision trees also implement rules Decision trees also implement the

decision strategies Sequence of rule firing is built in RBS does not have the sequence

built in, the conflict resolution strategies are developed.

1 rbs cs 331/531 dr m m awais production systems symbolic based rule based system based on simple...

Documents