Lecture 33

IIT Kanpur July 201
20 followers
1 Video View·Mar 1, 2024
[Music]
uh welcome back my dear friends
participants uh this is the
multicriteria decision making course
under NP Muk and my good name is runand
s Gupta from the IM Department at I khur
in India so this is the 33rd lecture out
of the 60 lecture course which is spread
over 12
weeks and each week we
have about uh five classes each class
being for half an hour now if you
remember the broader umbrella or the
concepts under which we were discussing
lot of Concepts and ideas is multier
decision
making and then we are
under going the understanding of
multiobjective decision making then
we'll consider multi-attribute decision
making and multiattribute utility Theory
and then proceed accordingly as was
discussed in the syllabus when we
started the class so first part is
quantitative then qualitative like
objective subjective concept more
mathematical modeling less of math
mathematical modeling now if you
remember in the 32nd class we were
discussing the problem where harvesting
has to take
place and the person has four different
crops to cultivate igate and my there
was one um point which I mentioned that
when you consider the concept of
revenues I'll come to the problem when
you consider the revenues you have one
set of objective function but the
constraints are same then you had the
cost function so consider separately
with the same objective functions then
you can analyze the problem from either
a net profit perspective or a net loss
perspective now the interesting part for
this problem was that they were both
negative and positive terms considering
the net profit on the net loss so the
problem would
basically if considered from the net
profit or net loss would have both the
positive and negative terms that means
producing or not producing would
basically either push up the profit or
lower down bring it more on the lower
side for the loss such that a balance
has to be made so considering that it is
an linear programming example for gold
programming and we'll consider
more such
problems but the problems will be
discussed uh not from the Practical
point of view that that means there's an
example it will be considered with a
very simple mathema model the reason
being it is in three dimension so it
will be easier for us to communicate how
the
overall the feasible space and how the
solutions would look
like and having said that if you
remember we had discussed in simple
concept how multi-objective problems can
be solved either you weit them with the
function
lambdas or you basically consider the
maximization or the minimization of only
one objective function and bring the
other objectives as a part and parcel of
the constraints so we'll highlight that
the same concept where we consider a
convex
combination of the objective
functions and also the considering the
case where you can
consider the one of the objective
function has to be optimized and the
rest can be passed on to the constraints
we'll consider for the nonlinear
programming also by by the word
nonlinear programming I mean the simple
concept of quadratic programming so to
understand the general concept of the
problem where there were four crops so
they were rice wheat Maze and bazra so
the problem
is the farmer
cultivates these
crops so the problem we have discussed
but I'm just mentioning selling prices
per kg are given as respectively as 28
25 30 and 24 the cost selling price
would be a more to for the revenue cost
per acre for the four crops are
respectively 33,200 30 ,500 35750 and
27,000 while the output per acre for the
same for these four crops
are 1250 that is 1,250 1,150 and 1,200
and finally being
1,100 and based on that we already have
the information that the total acreage
is 1,000 2% of that cannot be
utilized and another two was obviously
the crop have to be produced in positive
quantity either zero or more than
zero and also there were other two
constraints depending on the the area
where the person is is growing the crops
so due to water shortages rice cannot
exceed as it was
mentioned rice cannot exceed that of the
combined acreage of Wheat and bajra and
wheat being wheat being the staple diet
hence his production should be not
should be more than 50% of the combined
production of rice bajra and Maze so we
had the revenue function which I'm
marking in in
green the cost function which I'm
marking in Red so either you optimize
the revenues or optimize the cost
optimize the revenue means increase it
optimize the cost means decrease it but
the constraints remain the same when I
consider the constraints those four
which were mentioned they are as follows
constrain one I'll use the dark red
color you can use 98% of the total
acreage so X1 X2 X3 X4 all added up have
to be less than equal to 980 so that X1
X2 X3 X4 as we have said and I kept
repeating is the acreage for each and
every crop rice wheat and bajra and
Maze similarly due to the production
constraints about uh we water
requirement for rice and wheat being the
staple
diet the second constraint I'm marking
in
blue and the third constraint I'm
marking in Violet and
obviously I've considered each of them
as less than 400 they in the problem we
had considered between 100 and
400 now
comes the formulation I would not uh
solve this problem for a reason because
this being in four dimension it will be
easy not very easy for us to visualize
so in order to visualize we'll take a
three-dimensional um problem with
variables as X1 X2 and
X3 now if I consider the maximization
case only so I'll put the maximization
case as in green so the maximization
position and the constraints which were
already discussed I'll mark them
with the dark red color as
one total
acreage two being for
rice and three being for staple di of
Wheat and fourth being less than
400 now if I consider
the minimization problem then it was
basically minimization of the cost now
what we we have we have seen that if you
only minimize the cost best would only
nothing else to be considered so
obviously it's best to produce
nothing so the minimum cost would be
zero now what we do is that the moment
we start producing the cost would be
more than zero step by step
so I'm not considering the sign cost in
the sense it is a negative flow of money
out of the pocket but I'm considering
the quantum wise so if I only
concentrate on
the cost the cost function was
this it was obviously greater than has
to be greater than zero in case we
moment we produce one unit of any of
this crop of this four the cost will be
uh negative but if I consider without
considering the sign it will be positive
so that's why I'm increasing the value
zero which is the minimum value with the
amount of Delta so Delta can be 10 20 30
40 I can choose the value accordingly
and find out the values of the
maximization function based on the fact
that the last constraint which is marked
in blue is true for different values of
Delta and here Delta is is positive now
which means that we are as there are
only two objective functions so when we
optimize one the other comes into the
constraint that means the number of
constraint increases by one now the
question would be can we formulate the
problem in the similar way where we try
to minimize the cost with the constraint
corresponding to the revenues coming to
the constraint answer is yes so here is
the
problem so we can minimize the cost so
it's minimization problem because cost
has to be
minimized and the corresponding
constraints are exactly the same as they
were in the last slide so this was one
related to the
acreage of 22% not being utilized this
is related to the wheat what utilization
was for the rice third one was because
wheat was a staple D and fourth one was
basically less than equal to 400 and if
I consider the constraint the added
constraint coming from the cost if I
solve the the coming from the revenues
if I solve the revenue
function the value for the revenue
function POS maximum value would
be this value as given here which is 21
lakhs now that was the best profit now
the moment I want to be basically
increase some production that means cost
will become negative when I'm minimizing
it obviously the amount of the profit
will decrease so a negative component
will be has to be added so that's why it
is minus Delta where Delta is positive
so if I look add this constraint added
one which is related to the cost
revenues considering we are trying to
minimize the cost it is less than equal
to the max value minus the Delta value
and for the previous problem when we are
trying to maximize the revenue based on
the one of the con extra constraint
being the cost which has to be greater
than equal to Zer because Z was the
minimum value greater than then it will
be a addition of this positive Quantum
Delta now how it would look like in a
simple problem which we can understand
let us solve
it now to make things a little I'm not
saying to make it difficult but to make
it much more comprehensive I have not
considered two objectives but I have
considered three objectives it's point
one point number two for Simplicity it's
not that the problem cannot be solved
for Simplicity I have taken all the
optimization being maximum so maximum of
the first
function which I Mark as
FS1 maximization of the second function
which I Mark as
FS2 and maximization of the third
function which I Mark as
fs3 so the
point of maximization could also have
been replaced by
minimization so this could have been
also minimization of first function
similarly it could also have been
minimization of the second function it
could also have been minimization of the
third
function
so the ID Remains the Same how we solve
it only greater than less than sign the
Delta value which I mentioned will be
adjusted accordingly what are the
constraints so it's a three dimensional
one we have to visualize and I'll draw
it the first one is the sum of X1 X2 X3
is less than equal to 30 and by the way
all X1 X2 X3 are all
positive the second one is X1 + X2 is
less than equal to
20 the third one is X1 + X3 is less than
equal to 20
the fourth one is X2 + X3 is less than
equal to 20 and obviously as I said I'm
repeating it X1 + X2 X3 are all positive
now we to start the problem we'll forget
that it's a multiobjective problem we'll
only concentrate on the single objective
single objective which is F1 single
objective which is F2 single objective
which is
F3 optimize them and then proceed so we
will consider maximization again
maximization could also have been
minimization so I'll again write it if
it is a repetion please excuse me but
I'm trying to highlight it accordingly
so the max could have been Min
also so here it is maximization of the
first function which is shown in Red by
the way why I mentioned red why I
mentioned Violet and why I mentioned
blue would become very clear as we solve
the problem so the transformation of the
objectives into the constraints step by
step for these three problems why three
I'll come to that later step by step as
we do so the coloring Cod coding will
make things much much easier for you to
all to understand and for me to
explain now why three problems because
if you consider there are three
objectives so technically we would
basically be considering the first
objective and transforming the other two
objectives into the constraint then take
the second objective and transform the
first and the third into constraint and
the last case would be take the third
objective and transform the first and
the second into the
constraints so the first objective I'll
denote the optimum value I'll denote by
F1 star and it's a red color C the
second objective which is in Violet so
when I call the coloring scheme you will
understand so the first objective just
for repetition I'm saying it is + 2 X1 +
X2 - X3 shown in red color the second
objective shown in violet color is - X1
+ 2 X2 + 3 XT and it's marked the
optimum value when you Optimum find out
the single objective F2 star again
Violet third objective shown in light
blue is plus X1 - X2 + 2
X3 and when we consider the four sets of
constraints so if you read it when you
consider the four sets of constraints
what are these
these are they are already there the
four sets are technically
these
1
2
3 4 and the last one which is related to
X1 X2 X3 being greater than zero is
true because they are all positive so
when you consider this four constraints
the objective functions solution for
separately for first second third are F1
star in red color shown F2 star in
violet color shown F3 star in blue color
shown now let us solve it it's a simple
linear programming and we solve it the
problem is maximization of the first
objective I'll denote as objective
one number
one the
constraints remain exactly the same as
they are so it's 1
[Music]
2 3 4 and obviously the fifth one is
this one so I'll write it
five when I solve it the solution point
is for
X1 the value is 20 for X2 is 0o X3 is
zero so the objective function when we
solve it it's 40 how does it come
it comes as let me use a darker color so
this is
X1 X2
X3 solution
is 20 into this is in multiplication
sign into
2 + 0 - 0 = 40 so this is the value of
40 now let us solve the second objective
separately again separate single
objective so when we solve
it the objective function is here what
I'm marking the five
constraints
are
here when we solve it the value of X1
which was 20 now zero X2 which was 20
now is which was Zero now it's 20
and XT was 0 is 0 so the object function
is min -
0 + 2 this this star is a multiplication
sign into 20 + 0 is 40 as shown here in
the violet color second problem also
solved as single objective let us solve
the third
problem third problem is given
here which I'm which I already repeated
the coloring scheme so is blue in color
which is minus X1 so this would
be uh one minute is plusus plus
plusus plusus plus plusus plus this
would be
plus so this is the objective function
the constraints are
again
1 2 3 4
5 when I solve it
X1 which was first was 20 then 0 is
still zero
X2 which was in the first problem
objective function one was Zero
objective function two solution was 20
now for objective function 3
zero and X3 which was zero in both the
cases when we Sol the object function
one object function two now it is for
objective function three it's 20 the
solution
is + 0 - 0 + 2 into this is the
multiplication sign into 20 =
40 as shown here by F3 star is equal 40
so this coloring scheme now makes sense
that we can
differentiate so the corner points in
one sense
20 0 0 02 0 0 0 20 are the corresponding
optimal points for objective Function
One shown in red then for object
function two shown in Violet and last
being for objective function three which
is shown in blue and the these object
functions are F1 FS2
fs3 now we will slowly
bring the constraints in um constraints
to be in a such a way
that the objective functions which are
not being considered to be optimized
will be considered as a part and parcel
of the
constraints so the problem is first
case is I'm taking the objective
function one so I'll write it obj number
one
and the constraints one is there
perfect constant two is there constant
three is there constant four is
there and constant five is also
there now the Violet and the blue one
are the corresponding objective
functions two and three being brought
into the
constraints so now is objective fun
maximization problem we have solved them
individually and we found out the values
individually for each F1 FS2 F3 which is
fub1 star F2 star F3 star was 40
respectively for all the three
cases so we will so generally if you
want to find out we'll try to basically
change the values of this
F1 FS2 F3 corresponding to the case when
they are in the constraint
and solve the problem accordingly now in
this case objective function two and
three are in the constraints so F2 star
violet F3 star blue should be decreased
from the value of 40 in order to see
that how that we can reach the values
corresponding to the case that we have
fixed the values of FS2 objective
function 2 fix the values of F3
objective function 3 and trying to
optimiz
the objective function one so what we
mean is this let me use the violet color
so this was actually 40 single so in
case if I subtract say for example 20
which means the objective function two
is being maintained at
20 and if objective function 3 which is
blue I'll use the blue color which
originally was 40 single
objective and I I I'm taking subtracting
30 which is
10 So based on the fact that the right
hand side for the last two constraints
this should be less than equal to
20 and this should be less than equal to
10 and based on that we will solve and
find out that what is the value of the
objective function
F1 which
means that once we have
fixed
FS2 at
20
F3 at
10 the value of fub1 which you will get
this whole set corresponding to the
values of FS1 FS2 F3 obviously X1 X2 X3
which is of the decision variables will
come later on will give us the best
optimum solution which is
required and we will forward it the
correspondingly problem
accordingly
now the issue is that this value of 20
and 10 which you have considered
considering that I have subtracted 40
from
the constraint second last one which is
related to the objective function two
and I have subtracted 30 from the
constraint last one which is
corresponding to objective number three
can they be changed answer is yes they
can be changed would we get different
results yes we will get different
results can they be plotted yes we can
plot them and we'll consider them
accordingly now we solve
it if you solve it if you remember there
were values of Delta 2 at Del Delta 3
which was corresponding to the second
last constant and the last constant
second last being for the objective
function two and third last constant
being of for objective function 3 so I
have also continued using the coloring
scheme as it is Delta 2 in
Violet Delta 3 in
blue and obviously the F1 start star why
I'm using a double star because F1 star
was the best value single objective 40
for each f140 F1 star 40 F2 star 40 F3
star 40 so we'll take fub1 double star
to denote different combinations of of
of the constraints corresponding to
Second objective third objective how
they affect F1 so I Mark take it in the
last column now I take a value of delta
2 22 Delta 3 22 which is what the left
hand side
actually would be basically 4 t - 22 so
that means less than equal
to 18 here also it will be so this is
equal to
18 This is 40 -
22 is equal to 18 similar 23 would give
me
17 this is
17 this is 15 this is
15 this is
15 this is 18
this is 18 if you look at this means the
first and second column I'm I'm I'm
proceeding this is equal to
15 this is
15 this is equal to
13
13 and
15 So based on that I will solve so the
first problem which I solved which I'll
mark in this one this is let me use a
different color this
is case one for maximization of
objective one with the object two and
objetive three in constants and less
than equal to respectively as 22 22 the
optimum value comes out at as 191
0 and the objective function F1 star is
39 then I take case
two
which is 23 23 that means 17 17 less
than equal to less than equal to for the
second last constraint third last
constraint the values come out to be for
X1 X2 X3 as 18.5 1.50 0 and the value is
38.5 for the objective function
similarly for the less than equal to
constraint for the second last and last
constants are given which I have told
I'm not going to repeat it are shown in
the First Column second column and the
corresponding values of x's are found
out X1 X2 X3 when X's means the vector
and the values of F1 stars are
calculated so having said that it will
mean that with the addition constraint
we have been able to solve the first
objective giving very interesting
results we can repeat that for objective
function two being the object in the
function to be optimized the objective
function three to be optimized but
having said that the other objectives
which are not being optimized would
automatically come in the constraints
we'll
continue discussing this problem for
objective function two then objective
function three draw the diagrams and
then go into the concept of trying to
understand in the quadratic programming
case which is the nonlinear programming
with this I'll will end this 33rd
lecture have a nice day and thank you
very
much
[Music]
[Music]
[Music]
