This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try
refreshing the page, (b) enabling javascript if it is disabled on your browser and,
finally, (c)
loading the
non-javascript version of this page
. We're sorry about the hassle.
This solution committed several mistakes, which I would like to highlight.
It didn't check the conditions under which AM-GM could be applied. In particular, it required that x, y, z > 0 , which were not guaranteed till the condition that "a, b, c > 0" was added in.
It didn't check that equality could exist throughout all the inequalities. In particular, it had to check that 2 y 2 = 4 x 2 and that 4 z 2 = 8 y 2 could be simultaneously satisfied. Yes, this is trivial to show that we could have z = 2 y = 2 x .
Finally, it didn't check that the conditions on a, b, c AND the restrictions on x,y,z could be satisfied. In particular, we need to show that
(
a
+
b
+
3
c
)
=
2
(
a
+
b
+
2
c
)
=
2
(
a
+
2
b
+
c
)
can be satisfied with positive reals.
Mark Hennings shows that this can be accompllished with
a
=
3
−
2
2
,
b
=
2
−
1
,
c
=
2
(and their multiples).
How can you justify that x,y,z are non-negative reals?
Log in to reply
Thanks. We have added the condition that a,b,c are positive, and so x,y,z are positive.
Note that the quantity we are minimizing is homogeneous in a , b , c , so we can assume that a + b + c = 1 . Since the quantity we are trying to minimize is a continuous function of a , b , c over the compact region a , b , c , ≥ 0 , a + b + c = 1 , there must be a minimum value!
We need to minimize X = 1 + b 1 − b + 2 c + 1 + c 4 b − 1 + 2 c 8 c = F ( b , c ) = 1 + b 2 ( 1 + c ) + 1 + c 4 b + 1 + 2 c 4 − 5 for a + b + c = 1 , so for b + c ≤ 1 . Solving for ∇ F = 0 we see that ∂ b ∂ F = − ( 1 + b ) 2 2 ( 1 + c ) + 1 + c 4 ∂ c ∂ F = 1 + b 2 − ( 1 + c ) 2 4 b − ( 1 + 2 c ) 2 8 The vanishing of ∂ b ∂ F tells us that ( 1 + c ) 2 = 2 ( 1 + b ) 2 , and hence 1 + c = 2 ( 1 + b ) . Putting this result into the requirement that ∂ c ∂ F = 0 tells us that ( 1 + 2 c ) 2 = 2 ( 1 + c ) 2 , and hence c = 1 / 2 . Thus b = 2 1 ( 2 − 1 ) , and hence a = 2 1 ( 3 − 2 2 ) . The value of X for these values of a , b , c is 1 2 2 − 1 7 .
I am not going into the details of determining the Hessian matrix at this point to show that we have a minimum for F ; nor am I going to show that the value of F on the boundary of the region b , c ≥ 0 , b + c ≤ 1 is greater than 1 2 2 − 1 7 ; this is left to the interested reader!
The end result is that the answer is correct, but for a much more complicated reason than originally assumed!
Calvin Sir , Mark Sir you guys are way out of my league. I am just a beginner as compared to you. Please forgive my mistakes.
Log in to reply
It is a common mistake to not check the initial conditions of the theorem before applying it. Bear this in mind, and avoid doing so in future :)
I have some questions:
Question 1 :
Since the quantity we are trying to minimize is a continuous function of a , b , c over the compact region a , b , c , ≥ 0 , a + b + c = 1 , there must be a minimum value!
Compact region? What is that? Are you referring to Compact space ? I'm not familiar with topology.
Question 2 :
Why did you only solve for ∂ b ∂ F = 0 and ∂ c ∂ F = 0 only? How about ∂ a ∂ F = 0 ? I thought we need to solve for all "partial derivatives = 0"?
Question 3 :
I am not going into the details of determining the Hessian matrix at this point to show that we have a minimum for F .
Second Derivative Test is the special case of Hessian Matrix when n = 2 right? Okay.
Question 4 :
The end result is that the answer is correct, but for a much more complicated reason than originally assumed!
How is it "much more complicated"? It looks like a standard "compute the max/min given a boundary condition", no? Am I missing something else? Or are you saying that it's complicated because the actual full working is very tedious?
Question 5 :
After setting a + b + c = 1 , I tried to solve it via Lagrange multipliers , and it got ridiculously tedious and I stop midway because it's just too long and I'm not sure if it's worth my time or not. How do I identify whether this technique actually works or not? Or better yet: How do I know that Hessian Matrix is more effective than Lagrange Multipliers?
Log in to reply
In Euclidean space, compact equals closed and bounded. Continuous functions on compact sets have and achieve their extrema.
You will note that I used homogeneity to remove a from the function, so there is no need to differentiate w.r.t. a . A function of 3 variables subject to one constraint is, generally, a function of 2 variables.
In one dimension, the Hessian matrix is just the second derivative. Studying the Hessian matrix is the multidimensional analogue of the second derivative test.
OK, not really that much more complicated, but more involved than the solution originally proposed by the author.
Lagrange multipliers will work, but may well yield overly complex equations that are hard to solve. I started by looking to use LM, but started simplifying the target function using the constraint, and then realised that we could eliminate the need for the constraint altogether. There is often some advantage in using the constraint to simplify the function before you start!
Log in to reply
Thank you for your response. Your replies are always very insightful!
Problem Loading...
Note Loading...
Set Loading...
Let ,
a + 2 b + c = x
a + b + 2 c = y
a + b + 3 c = z
Solving for a,b,c we get
c = z − y
b = x + z − 2 y
a = 5 y − x − 3 z
x 5 y − x − 3 z + 3 z − 3 y + y 4 ( x + z − 2 y ) − z 8 ( z − y )
x 2 y − 1 + y 4 x + y 4 z − 8 − 8 + z 8 y
− 1 7 + ( x 2 y + y 4 x + y 4 z + z 8 y )
Applying AM. GM. Inequality to first two fractions
2 x 2 y + y 4 x ≥ 8
x 2 y + y 4 x ≥ 4 2 − ( i )
Applying AM. GM. Inequality to last two fractions
2 y 4 z + z 8 y ≥ 3 2
x 2 y + y 4 x ≥ 4 2 − ( i i )
Minimum value of the expression = -17 + (i) + (ii)
= − 1 7 + 1 2 2
a + b = − 1 7 + 1 2
− 5