0% found this document useful (0 votes)
152 views21 pages

Fuzzy Logic Final PPT 1-21

Fuzzy inference systems use fuzzy logic to map inputs to outputs based on predefined rules. They are useful for problems involving classification and pattern recognition. Fuzzy systems derive their name from the uncertainty associated with their functioning. They are entirely rule-driven, with inputs and outputs mapped according to rules specified during design. Fuzzy logic allows for partial set membership where an input can belong to multiple classes to varying degrees. This helps fuzzy systems better model real-world problems compared to traditional non-fuzzy logic.

Uploaded by

abhi abhi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
152 views21 pages

Fuzzy Logic Final PPT 1-21

Fuzzy inference systems use fuzzy logic to map inputs to outputs based on predefined rules. They are useful for problems involving classification and pattern recognition. Fuzzy systems derive their name from the uncertainty associated with their functioning. They are entirely rule-driven, with inputs and outputs mapped according to rules specified during design. Fuzzy logic allows for partial set membership where an input can belong to multiple classes to varying degrees. This helps fuzzy systems better model real-world problems compared to traditional non-fuzzy logic.

Uploaded by

abhi abhi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 21

Fuzzy Inference Systems

 
Fuzzy Systems  

•  With  fuzzy  logic,  the  system  is  made  to  give  the  most  probable  output  
to  any  kind  of  input  based  on  the  predefined  rules.  

•  Other  interes=ng  use  of  fuzzy  systems  is  in  classifica=on  and  pa>ern  
recogni=on  problems,  where  they  are  able  to  easily  determine  the  
output  class  that  the  input  corresponds  to.  

•  Fuzzy  systems  get  their  name  from  the  uncertainty  or  probability  they  
associate  with  the  various  stages  of  func=oning  as  they  calculate  the  
outputs  from  the  applied  inputs.  

•  Fuzzy  systems  are  en=rely  rule  driven.  Mapping  of  the  inputs  and  
outputs  is  accomplished  by  the  rules,  which  are  specified  during  the  
design.  
Historical Note  

•  The  history  of  fuzzy  logic  goes  back  to  the  days  of  Aristotle  and  the  
binary  logic  represen=ng  true  and  false,  which  began  the  development  
of  logic  in  the  history  of  humankind.    

•  Mul=  logic  also  evolved  about  the  same  =me,  but  not  to  a  very  good  
extent.  

•  The  field  a>racted  the  a>en=on  of  numerous  researchers  worldwide  


and  ini=ated  a  great  deal  of  work  in  this  field.  

•   Fuzzy  logic  then  joined  the  applica=on  domain,  where  it  has  been  
used  in  numerous  systems  and  consumer  applica=ons  based  on  
Intelligent  control.  
   
Fuzzy Logic  

           Logic  
•  Every  mapping  of  the  inputs  to  the  outputs  is  done  using  a  set  of  
guidelines,  or  func=ons,  that  are  the  inherent  proper=es  of  the  system  
being  considered.  

•  This  knowledge  removes  the  gap  between  human  and  machine  


understanding.  

•  We  also  defined  knowledge  as  “a  func=on  that  maps  a  domain  of  
clauses  onto  a  range  of  clauses.  

•   The  func=on  may  take  algebraic  or  rela=onal  form  depending  on  the  
type  of  intelligent  applica=ons  
Logic  

•  Consider  the  rule:  


•  If  (X  marks  are  more  than  80)  &  (X  a7endance  is  more  than  75%),  then  
(X  grade  is  A).  
•  Here  the  if  part  states  all  the  condiBons  that  if  true  lead  to  the  acBon.  
Problems with Non fuzzy Logic  

•  This  means  that  either  the  ac=on  will  take  place  or  it  will  not.  If  the  
condi=on  were  the  set  of  condi=ons  joined  by  logical  operators,  then  
the  same  concept  holds  true.    
•  Once  again,  the  various  condi=ons  are  evaluated  using  the  state  of  the  
system  and  are  worked  using  the  logical  operators.  

•  In  the  real  world,  however,  this  might  not  give  a  very  realis=c  picture  
of  the  en=re  system.  Consider  this  example:  

•  If  (driver  experience  is  high)  &  (road  is  bad),  then  (accident  risk  is  
moderate).  
•  If  (driver  experience  is  low)  &  (road  is  bad),  then  (accident  risk  is  high).  
•  If  (road  is  good),  then  (accident  risk  is  low).  
 
Problems with Non fuzzy Logic
Problems with Non fuzzy Logic

•  Working  with  these  rules,  it  may  easily  be  seen  that  for  a  person  with  5  
years  of  experience  driving    on  a  road  with  road  index  0.5,  the  accident  
probability  is  moderate  (or  0.4).  
Problems with Non fuzzy Logic
Fuzzy Logic

•  In  fuzzy  logic,  every  input  belongs  to  every  class.  The  degree  of  
associa=on  of  the  input  to  the  various  classes  varies.  
•  This  associa=on  may  be  very  strong  to  some  class  but  and  weak  for  
other  classes,  or  the  associa=on  may  be  moderate  for  all  classes.  
•  In  our  example  of  the  road,  we  find  that  under  fuzzy  logic,  experience  
can  be  high  and  low  at  the  same  =me.  
•   Thus  the  driver’s  experience  may  be  high  to  the  extent  of  80  percent  
and  low  to  the  extent  of  10  percent.    

•  This  means  the  driver’s  behavior  closely  follows  the  behavior  of  
experienced  drivers,  but  the  10  percent  associa=on  indicates  that  this  
behavior  to  some  extent  follows  the  behavior  of  inexperienced  drivers.  

•  This  gives  good  results  when  applied  over  real-­‐life  cases.  


When Not to Use Fuzzy  

•  Fuzzy  logic  follows  simple  English  rules  that  must  be  known  for  a  
system  to  have  an  effec=ve  intelligent  design.    

•  In  the  absence  of  these  rules,  the  performance  might  be  poor,  or  we  
may  have  to  apply  many  efforts  to  study  the  pa>erns  of  inputs  and  
outputs  in  search  of  the  rules.  

•  If  we  are  to  find  the  speed  of  a  vehicle  at  =me  t  and  accelera=on  a,  it  
would  be  be7er  to  apply  the  standard  mathemaBcal  equa=on  rather  
than  fuzzy  logic.    

•  Although  in  the  same  problem,  if  we  introduce  addi=onal  constraints  
that  mathema=cs  finds  it  very  difficult  to  cater,  the  problem  may  
become  fuzzy.  
Fuzzy Sets

•  We  have  already  discussed  the  concept  of  degree  of  membership  and  
various  classes  in  terms  of  fuzzy  logic.            
•  This  may  be  any  value  greater  than  or  equal  to  0  and  may  be  
represented  by  

•  This  is  shown  as  a/b,  where  a  denotes  the  element  of  the  set  and  b  
denotes  the  degree  of  membership  of  a  in  the  set.        

               where  we  assume  that  the  degree  of  membership  of  z  in  the  set  is  
given  by  the  funcBon  m(z).    
           Thus  it  is  natural  that  the  degree  of  membership  will  increase  as  z  
increases,  because  as  experience  increases,  the  driver  will  more  closely  
follow  the  characteris=cs  of  an  experienced  driver.  
Membership Functions

•  Every  element  is  denoted  with  a  certain  degree  of  associa=on  that  is  
given  by  a  func=on  known  as  the  membership  func=on  (MF).  
•  The  func=on  may  be  denoted  by  m(z),  where  z  is  the  element.  
•  Any  input  may  have  one  or  more  membership  func=ons  associated  
with  it.  
•  The  member  func=ons  for  the  two  classes  of  experience—low  and  
high—are  as  given  below  
Membership Functions

•  MFs  are  defined  by  the  system  designer  according  to  the  problem.  
Normally  designers  prefer  to  use  standard  membership  func=ons,  
which  have  been  used  in  numerous  problems.    
•  We  now  discuss  a  few  of  these  membership  func=ons.  
Gaussian Membership Functions  

•  This  widely  used  membership  func=on  denotes  either  a  sharp  


Gaussian  decrease  or  a  sharp  Gaussian  increase  in  the  membership  
value.  
Gaussian Membership Functions
Triangular Membership Function  

•  This  func=on,  which  denotes  a  straight-­‐line  decrease  or  increase  in  the  
membership  value,  is  used  in  situa=ons  where  there  is  a  simple  linear  
degrada=on  or  up-­‐grada=on  of  the  membership  value.  

•  From  that  point,  the  membership  value  starts  increasing  and  touches  a  
maximum  of  1  when  it  is  at  point  b.  
•   It  then  starts  decreasing  unBl  it  reaches  0  at  point  c.  From  c  onward  
the  membership  value  is  0.  
Sigmoidal Membership Function  

•  The  sigmoidal  MF,  which  depicts  the  sigmoidal  func=on,  is  given  by  
Other Membership Functions  

•  Commonly  Used  Membership  Func7ons  


Other Membership Functions

           
Other Membership Functions

 
 
 
 
 
 
 
 
 
 
 
 

You might also like