Professional Documents
Culture Documents
Markov Modeling
Pumping units in Two states!
Stochastic Modeling
P0(t) P1(t)
z P0(t)= { / ( + )} + { /( + )} e -( + ) t }
z P1(t) = { / ( + )} - { /( + )} e -( + ) t}}
z Case 1 Po *(t) : ( + )
Using approximate series and solving:
Po*(t) ={ /( + )}+{ /( + )} e -( + ) t }
Where: = 1-{/( + ) },
= 1+ {/( + ) } [ - ( + ) t]
What Was the Outcome?
z Good Outcome: Awarded PhD degree in Engineering.
z Not so Good Outcome: The work has very little practical
relevance.
z Are there any simpler techniques in decision research
which could model aleotarial uncertainty (due to
randomness in parametric data) and epistemic
uncertainty (due to perceptions that are invariably
vague, and ambiguity)?
z Yes. What is needed is Human Level Machine
Intelligence (HLMI). Fuzzy logic via CW could be one
such method of uncertainty analysis.
Broad Issues for Discussion
Hypothesis
Nature (Truth)
Design
Experiment Window
Deduction
Application Areas
z Medical Informatics
z Safety and Reliability Engineering
z Environmental Engineering
z Nuclear Management
z Bioinformatics
z Law ( Environmental , Land record, Constitution, and
Nuclear)-Ongoing
List of Selected Papers
1. Medical Informatics
3. Environmental Engineering
4. Nuclear Management
Ashok Deshpande and Vidyottama Jain Computing With
Words on Energy Options? Towards Decision Making
under Risk, International Journal of Nuclear
Management Dec 2010
Vidyottama Jain and Ashok Deshpande, Prospect Theory
on Energy Options?-Towards Decision Under Risk
ICRESH Dec 2010
Vidyottama Jain and Ashok Deshpande, Decision Making
undr Risk and Uncertainty CSI Communications
5. Bioinformatics
Efficacy of Fuzzy Polynucleoide Space in Phylogenetic
Tree Reconstruction
Cab Driver ProblemThe Point of Departure
Preview
Why?
Computing With Words
CW = NL-computation
Important Note:
Fuzzy logic is based on Computing With Words (CWW) but
Computing With Words may not necessary need Fuzzy logic
formulation. CW methodology can be developed using fuzzy
set operations and the decision maker can still arrive at
useful conclusion.
Levels of Complexity in CW
z Principal levels of complexity in CW: Level 1 CW and
Level 2 CW. Historically, Level 2 CW may be viewed as a
sequel to and a generalization of Level 1 CW.
z Today, it is Level 1 CW that is in preponderant use and a
center of research activity. In coming years, the intrinsic
importance of Level 2 CW is likely to gain recognition,
leading to a wide variety of applications.
z The concepts of a linguistic variable and fuzzy if-then
rule (Zadeh 1973) may be called the language, LNL, of
linguistic variables and fuzzy if-then rules. LNL is a
sublanguage of natural language, NL.
z In Level 1 CW and Level 2 CW, the universes of
discourse are LNL and NL, respectively.
Evolution of CW
CW
Level 1 Level 2
Evolution
universe of universe of
discourse discourse
LNL NL
explicit implicit
constraints constraints
Level 2 CW: Natural Language (NL)
Most Africans are tall
Most tall Africans are black
______________________________
Most2 Africans
are black
NL Level 2 CW
Linguistic Natural
Language (LNL) (LNL NL) is
Fuzzy Relation
Pressure is
High
LNL
If pressure is Level 1 CW
High then
volume is Low
Levels of Complexity in CW
Simple Examples
z Level 0 CW: Computing with Numbers
z Kalyani is 24
z Aashutosh is 2 years younger than Kalyani
________________________________________
z Aashutosh is (24 -2) years old
z Level 1 CW:
z Kalyani is young
z Aashutosh is a few years younger than Kalyani
______________________________________
z Aashutosh is (young-few) years young
Levels of Complexity:
Simple Examples
______________________________
z Most2 Africans are black
meaning
v-imprecisiation
1 1
0 v-precisiation 0
a x a x
M (meaning)-precise m-precise
V (value)-precise v-imprecise
m-precise m-precise
V-Precisiation
X: variable A
g-precisiation
X
a
s-precisiation
s-precisiation
400% (a)
Price of Onion in
Very Very High
Indian market (x) g-precisiation
s-precisiation is used routinely in scientific theories and
especially in probability theory
Defuzzification may be viewed as an instance of s-
precisiation
Professor Ashok Deshpandes Pravachan at ISI Bhubaneshwar
Workshop on Jan 15,2011- say around 11:45 hours
0 xa
(x-a) /(b-a) axb
= (c- x)/ (c- b) b x c
0 x c
1.0
f (x) = 0/11:00 + 0.2/11:15+ 1/11:45 +
0.7/12:00 + 0.45/12:15 +0/12:30
-kevel cut A principal objective of
0.20 human level
intelligence is
mechanization of these
0 remarkable human
capabilities.
11:00 11:15 11:45 12:00 12:15 12:30
Mechanization of these
capabilities is beyond
0 a b c the reach of methods
x based on classical,
l Support of a fuzzy set l Aristotelian, bivalent
logic and bivalent-logic-
Or mh precisiation based probability
theory.
Or co intensive definition
Two remarkable human capabilities: First, the capability to perform a wide variety
of physical and mental tasks, And second, the capability to reason, converse and make
rational decisions in an environment of imprecision, uncertainty, incompleteness of
information, partiality of truth and partiality of possibility.
Principal Rationals for CW
z Basic premises:
z Words are less precise than numbers
z Rationale A.
Use of Words
A B C
True Numbers
are known.
There is a
True Numbers tolerance for
Linguistic
not known Precision.
Summarization
Words are
good enough
1.0 Crisp or Classical Set
A Child is Born
Life
Is this OK?- NO
0 No Life
0 9
Fuzzy Set
Age in Months 1.0
This is
OK
Which set could
be a real
representation?
0
0 4 9
Age in months
Why Fuzzy Sets are Different
from Crisp Sets ?
z De Morgans Laws A A
(Laws of contradiction)
1
1
A
A
0 0
x x
[a] [b]
0
x
[C]
Fuzzy A 1
1
FuzzyA
0 0
x [b] x
[a]
0 x
[c]
Excluded Middle Laws for Fuzzy Sets [a] Fuzzy Set A
and its compliment; [b] Fuzzy A A X ( law of
excluded middle); [C] Fuzzy A A Null set ( law of
contradiction)
Linguistic Summarization Via
Granulation of A Function
Y f granule
L
M
S
0
0
Y S M L
medium medium
*f (fuzzy graph) perception
fsummarization*f :
If X is small Then Y is small
If X is medium Then Y is medium
X If X is large Then Y is small
0
49 /88
Fuzzy Logic Gambit
z Fuzzy Logic Gambit = deliberate imprecisiation through
granulation, followed by graduation
How?
Point of Departure
z The point of departure in CW is a question, q, of the
form: What is the value of a variable, X. Associated with
q is a question-relevant information set, I, expressed as
X is i, meaning that an answer to q, Ans(q/I), is to be
deduced (computed) from I.
z Typically, I consists of a collection of propositions, pi,
I=1, , n, are expressed in a natural language. I is said
to be linguistic if it contains propositions which are
drawn from a natural language.
z The pi are carriers of information about X.
z In general, some of the pi may be drawn from external
sources of information, typically from world knowledge,
+pi. Computation of Answer (q/I) is described in part by
an aggregation function, f.
Basic Structure of CW
NL q
CW
P
I Engine Ans q/I
P+
NL
NL
ExampleDurgas Age
z Question, q: What is age?
z Information set, I:
z p1: Durga has a son in mid-twenties.
z p2: Durga has a daughter in mid-thirties. Raw
z +p3(world knowledge): Usually mothers Information
age at birth of child is between Unprecisiated
approximately 20, *20, and
approximately 40, *40.
ExampleDurgas Age
Degree
Perception of Degree
(Fuzzy)
NL CW
natural
precisiation computation
language
description
FL
fuzzy logic
Paradigm Shift
CWW on Indias Energy
Options?-
Towards Decision Making
under Risk and Uncertainty
V(xi , pi ) = v(xi ) w( pi )
Where pi is perceived probability of outcome xi w (pi )
is the probability weighting function and v (xi) is value
function.
Prospect Theory
p
w( p ) =
p + (1 p )
CCW-2
FL
CW-1
Energy Options for India in 2030
with Gain and Risk Scenario
z Projected total energy needs in the year 2030 is
1,000,000 MW.
z Scenario 1:
Expected Power generation (gain) through hydro
energy is medium (as the water resources are
diminishing due to environmental degradation) on the
basis of water availability. The probability of its failure
i.e. risk due to accident (man made/ natural disaster)
is also assumed to be medium by the domain expert
group.
z Scenario 2
Computed energy generation (gain) using coal and
oil/& gas termed as thermal power is very low. Also
its risk of failure is very low.
Energy Options for India in 2030 with Gain and Risk Scenario
z Scenario 3
A(x)
0
x
0
A(x ) = (x-a)/ (b-a )
0 x 100,00
100,000 x 110,000
0 0 x 10000
1 x 11,000
A(x ) = (x-a)/ (b-a ) 100,000 x 110,000
Note: m stands from machine and h for human.
1 x 110000
0 0 x 110,000
The mathematical B(y )= ( x-a)/ (b-a ) 110,000 x 160,000
functions of membership 1 x 160,000
{A (x), B (y) and C (z)}
0 0 x 1000
functions for gains, total
= (x-a)/ (b-a ) 1000 x 14,000
energy requirement, and C(z ) 1 14,000
losses (expressed as
energy equivalent) are as
follows: X=gains, y=Requirement, z=Losses
What Next?-----
Max-Min Composition
Physical Analogy
Max-Min Composition
Link 2
Tensile Force
Link 3
.8 .3 .4 .4 .5 .5 .6 .7 .8 .8 .1 .8 .5 .5 .5 .4 .7 .8 .9 .8 .9
T= Ro S
where : R is a crisp relation on the Cartesian space X Y
and S is a crisp relation on the Cartesian Y Z space,
and T is a fuzzy relation on the X Z space.
T ( x, z)= (R ( x, y) S ( y, z))
yY
Where T is the characteristic function of T in the
interval [0,1]. This function measures the strength of
the relation, i. e., a value of 1means full relation and 0
no relation. is minimum and, is a maximum value.
Fuzzy Relations
Fuzzy max-min Composition
x2 0.08 0.08
Hy
x 2 0.5 0.5
T
~ ~
z 2200
= = 0 . 122
For thermal:
2
x 2 18000
z1 z z z
For hydro: = 0.16, 1 = 0.133, 2 = 0.18, 2 = 0.15
x1 x2 x1 x2
For Nuclear: z1 z z z
= 0.11, 1 = 0.1, 2 = 0.122, 2 = 0.11
x1 x2 x1 x2
Conclusion
Ration of Loss /gain is minimum in case of Nuclear
energy option. This is the reason for deciding
Nuclear Energy as the viable option or prospect.
Philosophy of Fuzzy Logic
NOT-A A And NOT-A A
Fuzzy Apple
A Non Traditional View
of Fuzzy Logic
Four Corner Stones of Fuzzy logic
graduation granulation
FUZZY LOGIC
precisiation
generalized
constraint
Rule base
Defuzzifier
Fuzzifier
(knowledge base)
output
input
Inference engine
Feedback Control System for Air Conditioner
Fuzzy logic Controller :Basis
z Theorems
1. FAT
Fuzzy Approximation Theorem has raised the
machine IQ of camcorders and car transmissions
and helicopter stabilizers and so on..
2. FAM
Fuzzy Associative Memory theorem deals with
firing of fuzzy rules
The FAT ideal has a simple geometry and
cover a curve with patches as shown below
Sensor Circuit
CO2 Temp
Sensor Sensor
Cooling
fan
Microcontroller
Blower fan
Controller Design
Controller selected-PIC16f73
RISC processor
8 bit in built ADC
3 I/O ports
2 PWM modules
119
Temperature sensor
Sensor selected LM35
120
CO2 Sensor
Sensor Selected-TGS4161
Range 350 ppm-5000ppm
Suitable for Indoor air Quality control
High selectivity to carbon dioxide
Compact size
Low dependency on humidity
Low power consumption
Long life and low cost
121
Proteus Simulation
Hardware Results
Final Circuit Diagram
Concluding Remarks
ZERO
Swami Vivekanand says:
x , x 0
v( x ) =
( x) , x < 0
Where, measure the curvature of the value function for
gains and losses, respectively, and is the coefficient of loss
aversion. This value function for gains and losses is
increasingly concave and convex respectively for, <1. The
weighting function, defined by Tversky and Kahneman [10], is
an inverse-S-shaped weighting function. It is concave near 0
and convex near 1 as presented in the Fig. 3. It is very clearly
explaining the fourfold pattern of risk attitudes as the low are
overweighed (leading to risk seeking for gains and risk
aversion for losses) and high probabilities are underweighted
the weighting function (leading to risk seeking for losses and
risk aversion for gains). It also satisfies Allais paradox.
Therefore, this modified inverse-S-shaped weighting is more
consistent with a range of empirical findings.
Figure 3. Progressive development in the area of decision makingh under Following Lattimore et al. [5], the weighting function can
risk and uncertainty. be parameterized in the following form
theory in their actual decision making behavior at least some
times. In response to their findings, Tversky and Kahneman
proposed a theory of choice, based on psychophysical model, p
which accurately describes how people go about making their
w( p) =
decisions. The Original Prospect Theory (OPT), suggested by p + (1 p)
Kahneman and Tversky in 1979, is based on non-linear
It assumes that the relation between w and p is linear in a log-
transformation of outcome and probabilities, which allow
describing psychological aspects of decision-making. The OPT odds metric. Here measures the elevation of the weighting
developed for simple prospects with monetary outcomes and function and measures its degree of curvature.
stated probabilities has three major characteristics: Observations on Cumulative Prospect Theory
OPT predicts that people go through two distinct stages
while taking decisions. In the first phase, decision makers are Invariably, the gains and losses are at the back of the mind
predicted to edit a complicated decision into a simpler of a decision maker while taking any decision. However, we
prospect, usually specified in terms of gains or losses. In the feel that he does not think of concave or steep convex
functions, while deciding options or prospects as suggested in availability. The probability of its failure i.e. risk due
CPT. In addition, estimation of the functions suggested in the to accident (man made/ natural disaster) is also
CPT model is difficult to achieve in practice. assumed to be medium by the domain expert group.
Computed energy generation using coal and oil/& gas
It is our belief that, in practice, most of the decisions are
termed as thermal power is very low. Also its risk of
taken by the domain knowledge experts with recourse to their
failure is very low.
perceptions. It would be desirable to try the formalism, which
requires less number of parameters in decision making under Consider Country A is in the final phase of its
risk and uncertainty, especially in reference to the energy development of nuclear fuel and also in agreement
options. with some other country for supply of nuclear fuel for
the sustained power generation. Therefore, power
IV COMPUTING WITH WORDS : CASE STUDIES generation through nuclear option is very high. Some
how, nuclear power plants have been over criticized
In this section, the application of CWW formalism has been because of one single unfortunate accident at
demonstrated for the two important issues of relevance under Chernobyl plant. Keeping in view the pubic
Indian context. The first issue deals with the planning of sentiments alone, risk of its failure is assumed as very
energy for the country till 2030. The Government of India is high.
making all out efforts to try out all the energy options and
4. A Word of caution: It is hoped that concerned
have recognised in research needs in the development of non
organizations will take cognizance of diligently
conventional energy sources such as bio fuel, solar energy, carried out environment impact assessment studies
wind energy and alike. In spite of the best efforts, the country and the public debate about the feasibility these plant
will have to use the conventional energy resources. How to locations. Only intellectuals, with clean image who
decide on these options? The write up will address this can work selflessly with the local community, should
question with possible solution. be involved in the process of site location for any
plant.
CASE STUDY 1
A] STUDY OF VIABLE CONVENTIONAL ENERGY Consider Fuzzy set A for generated power wherein
outcome is as gain, X, which depends upon the availability
OPTION FOR INDIA USING CWW METHODOLOGY
and over all cost / MWe. Fuzzy set B refers to the over all
DECISIONMAKING UNDER RISK
energy requirement, Y, while set C expresses the failure of
power (outcome as loss), Z, termed equivalent. The
Nature's real estate such as water, minerals, oil & gas,
membership functions for X, Y and Z are considered as
and coal -is the precious gift available for the welfare of
linear increasing functions.
mankind since time immemorial. In spite of the efforts
initiated by many countries in the development of non- Note: Understanding of meaning is a prerequisite to
conventional energy resources, it is too early to visualize that precisiation of meaning. Precisiation of meaning is a
the gap between supply and demand of energy for the peaceful prerequisite to computation; The expressions given below
purpose could be bridged in the near future at least in Indian justify the concept of precisiation proposed by Professor
context.. It is true that major decisions such as selecting Zadeh in his seminal work on CW methodology. It is of
energy options for the country are made without advance interest to note that the concept of precisiation, in the sense
knowledge of their consequences. Expected utility theory, in which it is used in CW, does not exist within linguistics
Prospect theory are some of the economic theories used in or computational linguistics. Expressions stated below
decision making under risk. explain the concept of precisiation.
z2 2200
z1 z2 For thermal; = = 0.122
x 2 18000
y1 0.6 0.6
SN =
~ y 2 0.69 0.77 For hydro;
z1 z z z
= 0.16, 1 = 0.133, 2 = 0.18, 2 = 0.15
x1 x2 x1 x2
Since fuzzy relation R N is defined from X to Y and fuzzy
~ For Nuclear;
N
relation S is defined from Y to Z, then fuzzy max-min
z1 z z z
~
N N
= 0.11, 1 = 0.1, 2 = 0.122, 2 = 0.11
composition between R and S results into the following x1 x2 x1 x2
~ ~
N We can draw useful conclusion form the computational
fuzzy relation matrix T as
~ procedure that the ration of loss and gain is minimum in case of
Nuclear energy option. This is the reason for deciding Nuclear
Energy as the viable option. or prospect.
z1 z2
x 0.69 0.77 B] case study 2
TN = 1
~ x 2 0.69 0.77 ENVIRONMETNAL FRIENDLY FUZZY LOGIC BASED AIR
CONDITIONER
2) Hydro energy option
Global warming and continuous emissions of green
house gases into the atmosphere are the two key issues of
immediate relevance. In addition, suffocating in a small closed nuclear energy is the viable option. It is important to note that
room with air conditioner on is a common scene. This is due in CW, only perception of the domain experts, who have (and
to accumulation of CO2 in the room. The Permissible CO2 we believe) adequate understanding of the several issues are
concentration could be a 350 ppm (max) for human being. For needed. Therefore, there is less computational bother as the
example, in a small close cubical if there are 6-8 people, the decision makers are not supposed to carry out extensive social
CO2 level could be as high as 500. ppm. This calls for the surveys for the estimation of several parameters, which might
design of a separate sensor in addition to one that controls the not have any practical relevance and could be at best be
desired temperature. considered as an intellectual exercise. The need for extensive
research in policy issues is recognized.
Based on the defined principles of fuzzy logic, the
College of Engineering Pune (COEP) team has designed a Fuzzy logic based environmental friendly and energy
prototype air conditioner which has now been patented. Two efficient air conditioner once again demonstrates the utility of
fuzzy inference systems (FIS) are, therefore, designed and Computing With Words.
implemented. This is an ongoing effort and will take some
more time to finalise the marketable product. Figure 4 depicts ACKNOWLEDGMENT
the outcome of the implementation
We are deeply indebted and would like to express our
immense gratitude towards Professor Lotfi A. Zadeh, the father
of fuzzy logic for his motivation, all helpful and insightful
Hardware Results
suggestions, and constant encouragement. Assistance from
Vidyottma Jain , CY Patil and Roshan Kshirsagar is gratefully
acknowledged.
REFERENCES
[1] Zadeh, L.A; Computing with WordsWhy? and How? Presented July
12, 2010 WORLDCOMP 201 Las Vegas, NV USA.
[2] V. Jain and A. Deshpande, Prospect Theory on Energy Options?
Towards Decision Making under Risk, Submitted ICRESH 2010.
[3] F. H. Knight, Risk, Uncertainty, and Profit, New York:Hart, Schaffner,
and Marx, 1921.
[4] D. Kahneman and A. Tversky, Prospect theory: an analysis of decision
under risk, Econometrica, vol. 4, pp 263 291, 1979. A. Tversky and
Figure 4 Environmental Friendly and energy Efficient Fuzzy D. Kahneman, Advances in prospect theory: cumulative representation
of uncertainty, J. Risk Uncertainty, vol. 5, no. 4, pp 297323, 1992.
logic based Air Conditioner.
[5] P.K. Lattimore, J.R. Baker, and A.D. Witte, The influence of
probability on risky choiceA parametric examination, J. Econ.
Behav. Organ., vol. 17, no. 3, pp 377400, 1992.
V DISCUSSION AND CONCLUDING REMARKS
[6] A.Deshpande and V. Jain Computing with Words on Energy Options?-
The two valued logic and probability theory is the basis of Towards Decision Making under Risk and Uncertainty Int. J. Nuclear
prospect theory. It is a deep-seated tradition in science to Knowledge Management( accepted for publication)
employ the conceptual structure of bivalent logic and .
probability theory as a basis for formulation of definitions and
concepts. What is widely unrecognized is that, in reality, most Ashok Deshpande, Chair Berkeley Initiative in Soft Computing (BISC)-
concepts are fuzzy rather than bivalent, and, in general, it is not Special interest Group (SIG)-Environment Management Systems (EMS) holds
a PhD degree in Engineering and Technology. He is an Adjunct Professor at
possible to formulate a co intensive definition of a fuzzy College of Engineering Pune (COEP) India. In the past, Dr. Ashok was
concept within the conceptual structure of bivalent logic and Deputy Director, National Environmental Engineering Research Institute
probability theory. It is, therefore, our endeavor to develop a (NEERI) /CSIR. He has over 4 decades of R&D experience and has over 100
formulation for decision making with imprecise probabilities in publications in the Journals of international repute/conferences. Dr Deshpande
fuzzy scenario, which might be useful and more realistic, in has contribution in various facets of environmental engineering with special
general, in the decision making under risk for the country reference to water supply engineering, risk analysis for chemical process
industry, and reliability engineering applications to variety of engineering
energy policies. systems. Though worked on complex mathematical techniques, in early 80s,
Decision-making problems under risk and uncertainty is he got interested in fuzzy logic applications initially to environment
management systems and later to several systems in science and technology.
always a great challenge to the decision researchers for many
decades as there is no definitive theory, which could help in
arriving at the desired outcome. Dr. Ashok was WHO Adviser, Common Wealth Science Council Resource
Scientist, World Bank Project Director for the studies on Probabilistic Risk
Case study 1 on energy options was approached using Assessment for Chemical Process industry. He also organized a Workshop as
Computing With Words methodology wherein gains, losses WHO Adviser in 5 countries on Unaccounted for Water Management and also
assisted Danish International Development Authority (DANIDA) as a Project
and total energy requirements are expressed in linguistic terms Advisor.
or hedges. Precisiation of meaning is the key point in CW. We
once again conclude or confirm using CW methodology that
More importantly, Professor Lotfi Zadeh, the founder of fuzzy logic, after In March 2008 Professor Deshpande organized a Training workshop on Fuzzy
listening to many seminar talks, asked Professor Ashok Deshpande to be the sets and Fuzzy logic with Applications at the University of Illinois Chicago
Chair of Berkeley (BISC)-(SIG)-(EMS). He is also guest faculty at the USA (sponsored by Vshwamitra Research Institute Chicago) and in
University of California Berkeley and visiting scientist at Lawrence Berkeley Kathmandu, Nepal in 2009. Dr. Deshpandes has a passion and mission to
National Laboratory, CA. In 2004, Dr. Ashok was invited by the International propagate the use of fuzzy logic nationally and internationally. At present, his
Atomic Energy Agency (IAEA) as an expert for organizing a Training all doctoral students work only on fuzzy logic related topics. Professor Ashok
Program at CDTN, Brazil on Fuzzy logic with applications and again in 2006. Deshpande is deeply interested in Indian classical music.