Exercise 4.23

Exercise 23: A real-valued function f defined in ( a , b ) is said to be convex if

f ( λx + ( 1 λx ) y ) λf ( x ) + ( 1 λ ) f ( y )

whenever a < x < b , a < y < b , 0 < λ < 1 . Prove that every convex function is continuous. Prove that every increasing convex function of a convex function is convex. If f is convex in ( a , b ) , and if a < s < t < u < b , show that

f ( t ) f ( s ) t s f ( u ) f ( s ) u s f ( u ) f ( t ) u t .

Answers

Letting a < s < t < u < b , then t = λu + ( 1 + λ ) s for λ = ( t s ) ( u s ) . Hence

f ( t ) λf ( u ) + ( 1 λ ) f ( s ) f ( t ) f ( s ) λ ( f ( u ) f ( s ) ) f ( t ) f ( s ) t s f ( u ) f ( s ) u s

Also, t = λ s + ( 1 λ ) u for λ = ( u t ) ( u s ) . Hence

f ( t ) λ f ( s ) + ( 1 λ ) f ( u ) λ ( f ( u ) f ( s ) ) f ( u ) f ( t ) f ( u ) f ( s ) u s f ( u ) f ( t ) u t

The first inequality can be rewritten f ( t ) f ( s ) = K ( t s ) for s < t < u , which shows that f is continuous from the right at s . Similarly, the second inequality can be rewritten f ( u ) f ( t ) = K ( u t ) for s < t < u , which shows that f is continuous from the left at u . Since s and u were arbitrary elements of ( a , b ) , we see that f is continuous on ( a , b ) .

Let f be a convex function defined in ( a , b ) and let g be an increasing convex function defined on the range of f . Then for a < x < b , a < y < b , 0 < λ < 0 , we have

g ( f ( λx + ( 1 λ ) y ) ) g ( λf ( x ) + ( 1 λ ) f ( y ) ) λg ( f ( x ) ) + ( 1 λ ) g ( f ( y ) ) .

Hence g f is convex on ( a , b ) .

User profile picture
2023-08-07 00:00
Comments

Proof. For this function f ,

( x z ) f ( y ) = ( z y ) f ( y ) + ( y x ) f ( y ) > ( z y ) f ( x ) + ( y x ) f ( x )

so

f ( y ) > z y z x f ( x ) + y x z x f ( x ) .

If λ = ( z y ) ( z x ) , it is evidently in the interval ( 0 , 1 ) . Also, λx + ( 1 λ ) z = y . This means

f ( λx + ( 1 λ ) z ) > λf ( x ) + ( 1 λ ) f ( z ) ,

which contradicts that f is convex. □

Lemma 1. For a convex function f over interval ( a , b ) , the function is monotonic for some 𝜖 > 0 on the interval ( x , x + 𝜖 ) for any x ( a , b ) .

Proof. Suppose not, and let the point x be the point that makes the value of f over this interval not monotone. From the previous proof, there must be some points a , b , c ( x , x + 𝜖 ) such that

f ( a ) f ( b ) < f ( c )

or

f ( a ) < f ( b ) f ( c ) .

Put 0 < 𝜖 < a x , there must exist a , b ( x , x + 𝜖 ) such that

f ( a ) f ( b ) and f ( c ) < f ( b ) ,

or

f ( a ) < f ( b ) and f ( c ) f ( b ) ,

so that f is not monotone in ( x , x + 𝜖 ) . However, this gives for b < a < b ,

f ( b ) f ( a ) and f ( a ) > f ( b ) ,

or

f ( b ) < f ( a ) and f ( a ) f ( b ) ,

which forms a contradiction. Thus, f must be monotonic on ( x , x + 𝜖 ) . □

Lemma 2. A convex function f over a monotone interval ( a , b ) cannot have any simple discontinuities, and thus, cannot have any discontinuities.

Proof. Assume that z produces a simple discontinuity at f ( z ) . First f ( z ) must be defined since otherwise

f ( λx + ( 1 λ ) y ) λf ( x ) + ( 1 λ ) f ( y )

would not hold when z = λx + ( 1 λ ) y , which means f ( p ) is defined on all p ( a , b ) . Since f ( z ) is a simple discontinuity, therefore, either f ( z ) f ( z ) or f ( z + ) f ( z ) .

Now assume, without a loss of generality, that f ( z ) > f ( z ) . Then, given 𝜖 > 0 that is arbitrarily close to 0 we can construct λf ( z 𝜖 ) + ( 1 λ ) f ( z ) . However, for f ( λx + ( 1 λ ) y ) , when z 𝜖 = λx + ( 1 λ ) y and 𝜖 0 , λ 0 ,

f ( λx + ( 1 λ ) y ) = f ( z 𝜖 ) = f ( z ) > f ( z ) = λf ( z 𝜖 ) + ( 1 λ ) f ( z ) ,

which contradicts the convexity of f . The other cases, for f ( z ) < f ( z ) , f ( z + ) > f ( z ) , f ( z + ) < f ( z ) follow by the same argument. Thus, monotonic subintervals of a convex function f must not have any simple discontinuities, and therefore, discontinuities in general since monotonic functions can only have simple discontinuities. □

Proof of ( a ) . Since the monontonic subintervals of f are continuous since f itself has no discontinuities, then f is continuous. □

Proof of ( b ) . Let f ( x ) , g ( x ) be increasing and convex. Then, for f ( x ) ,

f ( λx + ( 1 λ ) y ) λf ( x ) + ( 1 λ ) f ( y ) .

Since g ( x ) is increasing, we can say

g ( f ( λx + ( 1 λ ) y ) ) g ( λf ( x ) + ( 1 λ ) f ( y ) ) .

Also, since g ( x ) itself is convex, we can say

g ( λf ( x ) + ( 1 λ ) g ( y ) ) λg ( f ( x ) ) + ( 1 λ ) g ( f ( y ) ) .

Therefore,

g ( f ( λx + ( 1 λ ) y ) ) λg ( f ( x ) ) + ( 1 λ ) g ( f ( y ) ) ,

and g ( f ( x ) ) is convex. □

Proof of ( c ) . For the first inequality, let λ = ( u t ) ( u s ) . Because of the convexity of f we can say that:

f ( t ) f ( s ) ( 1 λ ) ( f ( u ) f ( s ) ) = t s u s ( f ( u ) f ( s ) )

so

f ( t ) f ( s ) t s f ( u ) f ( s ) u s ,

and we have the first inequality. Then, for the second one let t = λs + ( 1 λ ) u . Then, from the convexity of f :

f ( t ) u t u s f ( s ) + ( 1 u t u s ) f ( u ) f ( t ) u t f ( s ) u s f ( u ) u s + f ( u ) u t f ( u ) f ( s ) u s f ( u ) f ( t ) u t ,

and we have the second inequality. □

User profile picture
2023-09-01 19:48
Comments