CSE 840: Computational Foundations of Artificial Intelligence Feburary 24, 2025

Derivatives and Riemann Integral

Instructor: Vishnu Boddeti Scribe: Cherilyn Wang

1 Derivatives (1-dimensional case)

Definition 1 Consider U ⊂ R an interval, f : U → R. The function is called differentiable at a ∈

U if:
f (a + h) − f (a)
f ′ (a) := lim
h→0 h
exists.

We often write
df
f′ =
dx

an : a + hn , hn → 0, hn > 0
"Derivative is the slope of a function at point a"
"Slope of linear approximation of a function at point a"
f (x) = f (a) + (x − a)b (b is the slope and the derivative at a)

The function is called differentiable if it is differentiable for all a ∈ U. It is continuously differentiable

if it is differentiable and the function f ′ : U → R, a 7→ f ′ (a) is continuous.

Higher derivatives: We can repeat the process of taking derivatives:

df ′′ df ′
f′ = ,f =
dx dx

Notation: f (n) denotes the n-th derivative (if exists).

11-1
1.1 Important Theorems

Theorem 2 Differentiable ⇒ continuous, but continuous ⇏ differentiable

Let f be differentiable at a, then there exists a constant Ca such that on a small ball around a, we
have |f (x) − f (a)| ⩽ Ca |x − a|. In particular, f is continuous at a (This is also the case globally: if
the function is differentiable everywhere, it is also continuous everywhere).

Theorem 3 Intermediate value theorem for derivatives

f ∈ C ′ ([a, b])(i.e. functions on [a,b] that are once continuously differentiable), then there exists ξ ∈
[a, b] such that f (b)−f
b−a
(a)
= f ′ (ξ)

Theorem 4 exchanging limits and derivatives fn : [a, b] → R, fn ∈ C ′ ([a, b]) if the limit f (x) :=
lim fn (x) exists ∀x ∈ [a, b] and the derivatives f ′ converge uniformly, then f is continuously dif-
n→∞
ferentiable and we have:

f ′ (x) = ( lim fn )′ (x) → first take limit of fn , we obtain f , then compute derivative
n→∞
(1)
= ( lim (fn′ ))(x) → first compute fn′ , then take the limit
n→∞

11-2
2 Riemann Integral

Consider a function f : [a, b] → R, assume that f is bounded (∃l, u ∈ R, ∀x ∈ [a, b] : l ⩽ f (x) ⩽ u)

Consider x0 , x1 , ..., xn with a = x0 < x1 < x2 ... < xn = b. These points introduce a partition of
[a, b] into n intervals.
Ik := [xk − 1, xk ]

Define mk := inf (f (Ik )), Mk := sup(f (Ik )) (exists since f is bounded).

Define the lower sum
n
X
s(f, {x0 , x1 , ...xn }) = |Ik | · mk (length of Ik = xk − xk−1 )
k=1

11-3
Define the upper sum
n
X
s(f, {x0 , x1 , ...xn }) = |Ik | · Mk
k=1

Now define
J∗ := sup (s(f, partition))
partitions

J ∗ := inf (s(f, partition))

partitions

We call f Riemann-integrable if J∗ = J ∗ . Then we denote

Z b
J∗ = J ∗ := f (t)dt
a

Theorem 5 • f : [a, b] → R, monotone ⇒ integrable (i.e. x1 < x2 ⇒ f (x1 ) < f (x2 ))

• f : [a, b] → R, continuous ⇒ integrable (true even if f is continuous everywhere except at
finitely many points)

Shortcomings:

• Many functions are not integrable


1 x∈Q
Dirichlet Function : f (x) =
0 everywhere

11-4
 For any interval Ik = [xk , xk+1 ], Mk = 1, mk = 0
Then J∗ < J ∗ (J∗ :|b − a| · 0, J ∗ :|b − a| · 1)
• One cannot prove theorems about exchanging "integral" with "limit":
Z Z
?
lim fn dt = lim fk dt
n→∞ n→∞

• Hard to extend to "other space" (e.g. spaces with no notion of ordering, higher dimensional)

Later in the course, we will see that Lebesque Integration, which is a more modern and more useful
form of integration, can overcome some of the shortcomings of Riemann Integration.

3 Fundamantal Theorem of Calculus

Derivatives are intuitively slopes of the function at a particular point x, while integration computes
the area under the curve. However, it is not very obvious how and why these two are related to each
other. The fundamental theorem of calculus tries to relate these two.

Theorem 6 f : [a, b] → R is (Riemann)-integrable and continuous at ξ ∈ [a, b]. Let c ∈ [a, b]. Then
the function Z x
F (x) := f (t)dt
c

is differentiable at ξ and F (ξ) = f (ξ). If f ∈ C([a, b]), then F ∈ C ′ ([a, b]) and F ′ (x) = f (x) for all
′

x ∈ [a, b]

Proof: This is the theorem’s proof. □

Theorem 7 F : [a, b] → R is continuously differentiable, then

Z b
F ′ (t)dt = F (b) − f (a)
a

11-5