Open solomonxie opened 6 years ago
Tags: #LogisticGrowth #LogisticGrowthModel #LogisticEquation #LogisticModel #LogisticRegression
This is a very famous example of Differential Equation, and has been applied to numerous of real life problems as a model.
It's originally a Population Model
created by Verhulst
, as studying the population's growth
.
Refer to lectures: ▶Khan academy, ▶MIT Gilbert Strang's, ▶The Organic Chemistry Tutor, ▶Krista King, ▶Bozeman Science
Refer to Khan academy: ▶Logistic models & differential equations (Part 1)
Let's let P(t)
as the population's size in term of time t
, and dP/dt
represents the Population's growth.
Mr. Malthus first introduced the exponential growth theory for the population by using a fairly simple equation:
Where P
is the "Population Size", t
is the "Time", r
is the "Growth Rate".
Mr. Verhulst enhanced the exponential growth theory of population
, as saying that the population's growth is NOT ALWAYS growing, but there is always a certain LIMIT or a Carrying Capacity
to the exponential growth.
And combining the exponential growth
with a limit
, it's then called the Logistic Growth
.
And the logistic growth got its equation:
Where P
is the "Population Size" (N is often used instead), t
is "Time", r
is the "Growth Rate", K
is the "Carrying Capacity".
And the (1 - P/K)
determines how close is the Population Size to the Limit K
, which means as the population gets closer and closer to the limit, the growth gets slower and slower.
"It explains how density dependent limiting factors eventually decrease the growth rate until a population reaches a Carrying Capacity ( K )."
Carrying Capacity means the "celling", the "limit", the "asymptote".
It's gonna use the method
Separable Equations
, which introduced theinitial condition
asP₀
in this case.
We could directly solve the Logistic Equation as solving differential equation to get the antiderivative
:
But we still have a constant C
in the antiderivative
, which required us to introduce an Initial Condition
to get rid of C
and get the specific function:
Solve:
dP/dt = r·P(1-P/K)
.dy/dt = 10·y(1-y/600)
.K = 600
, which is the limit, the Carrying capacity.
Solve:
S(c)
.S'(c) = dS/dc = 0
S = 0 or 20,000,000
(0+20,000,000)/2 = 10,000,000
.▶ Refer to Khan academy: Worked example: range of solution curve from slope field
Solve:
Solve: Hint: Try a point or points in each quadrant, like Q1: (2,2), Q2: (-2,2), Q3: (-2,-2), Q4: (2,-2)
(2, 2)
, we will get y' = 0
, which means the line in Q1 would be horizontal.Solve:
6
is the initial condition, so we make 6
as a boundary.(0,6)
is a negative slope, which seems keeps negative until 4
.(4, 6]
.Solve:
(0, 1)
.x=0
value to get each slope, and eyeball it.Euler's method means an approximation by writing down every critical value in a table, and iterate many many times until it get closer to the target value.
Approximation:
Iterate table:
Solve:
Euler's Method
, we need to figure out how to get each column value, and iterate every row.Initial Condition
, so for the initial row, We know the x=-1, y=3
x & y
change, and they change differently.x
is from -1 to 2 in 3 steps, so Δx = (2 - -1)/3 = 1
Δy
. We know dy/dx ≃ Δy/Δx
, so Δy ≃ dy/dx · Δx
.dy/dx = (-1) - (3) - 2 = -6
Δy = dy/dx · Δx = -6 × 1 = -6
x = -1 +(1) = 0
and y = 3 +(-6) = -3
dy/dx = x - y - 2 = 0 - (-3) -2 = 1
Δy = dy/dx · Δx = 1 × 1 = 1
[Refer to Khan academy: Area under rate function gives the net change](Area under rate function gives the net change)
We're used to the function graph that using ordinary X-Y axes
. But it's also very often we use the X-R axes
(where Rate of the function as the vertical axis) for real problems.
Assume there's a function of a car's distance: D(t)
, and the car's speed is represented as a function r(t)
.
Instead of showing the function graph of D(t)
, we're showing the Rate function's graph:
Based on the distance formula Distance = speed × time
, we could know that D(t) = r(t) × Δt
By using a more calculus based term: The distance it traveled in a period of time is D(t) = ʃ r(t) dt
So in the graph above, the AREA of rate function
actually means THE DISTANCE TRAVELED IN A PERIOD OF TIME, which means:
THE DISTANCE IS THE ACCUMULATED SPEED.
Try to intuitively understand this in mind, then problems will be solved easily.
Strategy:
Rate function graph
.Solve:
What is the temperature.
This means we are looking for an actual value.initial condition
and a definite integral
:
Solve:
Solve:
►Jump to Khan academy for some practice: Motion problems (with integrals)
Displacement
literally means "the change in position", but actually it means the SHORTCUT of two points, the shortest distance between two points.
▶Jump over to previous note in Linear Algebra: Displacement is a vector, distance is a scalar.
Even if you've been travelling all the time without stops,
but your DISPLACEMENT
still can be 0:
Solve:
Solve:
Solve:
s(t), v(t) and a(t)
:
v(t) = s'(t)
or s(t) = ʃ v(t) dt
a(t) = v'(t)
or v(t) = ʃ a(t) dt
a(t) = 1
, so v(t) = ʃ a(t) dt = ʃ 1 dt = t + C
.v(3) = -3 = t + C = 3+C
, so C=-6
which makes v(t) = t - 6
s(t) = ʃ v(t) dt = ʃ (t-6) dt = 0.5t² - 6t + C
s(2) = 0.5*2² - 6*2 + C = -10
, so C=0
, which makes s(t) = 0.5t² - 6t
s(4) = 0.5 * 4² - 6*4 = -16
►Jump to Khan academy for some practice: Planar motion (with integrals)
Solve:
►Jump to Khan academy for some practice: Curve areas
Strategy:
ʃ f(x) dx
over the interval.Solve:
0 & x when f(x)=0
.f(x) = 0 = 2 + 2cos(x)
-> cos(x)=-1
-> x = arccos(-1) = π
[0, π]
.ʃ (2+2cosx) dx
over the interval [0, π]
2π
from the definite integral .Strategy:
y
.x
, we need to integrate in term of y
.ʃ f(y) dy
Solve:
f(y) = 15/y
y
:
Solve:
y
:
Strategy:
ʃ |f(x)-g(x)| dx
Solve:
Solve:
Solve:
f(y) - g(y)
to get:
Calculating area for polar curves
, means we're now under the Polar Coordinate
to do integration.
And instead of using rectangles
to calculate the area, we are to use triangles
to integrate the area for a curve.
There're a few notable differences for calculating Area of Polar Curves
:
Polar Coordinate
.Circle Sectors
with infinite small angles to integral the area.RAY
or two RAYS
from the origin.▶ Practice at Khan academy: Area bounded by polar curves
Refer to Khan Academy: Area bounded by polar curves
The most tricky part in Polar system, is finding the right boundaries for
θ
, and it will be the first step for polar integral as well.
Refer to youtube: Finding Area In Polar Coordinates
Why is this?
Better to try out on Desmos.com, to see if the interval produces the right shape.
Find out the boundaries of θ for integrating the shaded area: Solve:
Find out the boundaries of θ for both of the polar curves: Solve:
θ = 0
and end at intersection θ = π/6
Solve:
r
starts drawing at 0, all the way down for a round and goes back to 0.r
goes from π
then takes a round back to 2π
.π and 2π
Find out the boundaries of θ for integrating the shaded area: Solve:
r = -1
ends at r = 0
.r
and end of r
:
r = -1
, from the equation we get θ = 0
r = 0
, from the equation we get θ = π/3
Find out the boundaries of θ for integrating the shaded area: Solve:
θ ∋ [0,π]
, it makes r
negative: sin(π)-1 = -1
and sin(0)-1 = -1
.r
starts at 0 and ends at 0, not negative value, r=0
:
π/6
and 5π/6
.Solve:
3sinθ = 1+sinθ
, to get θ = π/6 and 5π/6
, which are the boundaries.Solve:
cosθ
shape, its boundaries are same with Quadrant-1: [0, π/2]
1+sinθ
shape, its boundaries are same with Quadrant-4: [3π/2, 2π]
Solve:
Solve:
►Jump to Khan academy for some practice: Arc Length. ▼Refer to Khan academy: Arc length intro
Solve:
f'(x) = 1/x
Refer to Khan academy: Arc length of polar curves
▶ Practice at Khan academy: Arc length of polar curves
Formula:
Find out the boundaries for each shape: Solve:
π/3
, through origin (0), ??
-π/3
and ends at π/3
.Find out the right boundaries for θ for the red arc length: Solve:
r
starts from 0 and ends at 2.r=2
the θ=0
.r=0
, there're many solutions, we need to consider which solution to use.r=0
and get:
θ=π/6
0, π/6
Solve:
r=0
and solve for θ
:
Parametric Curves
are from Parametric Equations
, means both x
and y
are functions, in terms of t: x(t)
and y(t)
.
►Jump to Khan academy for some practice: Arc Length.
Refer to xaktly: Parametric Equations ▼Refer to Khan academy: Parametric curve arc length
Solve:
Strategy:
Anyways, the key to solve these problems, is to HAVE STRONG SENSE OF 3D SHAPES. So that you could sense what is this problem asking for. And once you know what it's saying, you could easily solve it.
►Jump to Khan academy for some practice: Volumes of solids of known cross-section
Solve:
cylinder
from top to bottom in infinite many discs (of course it's rectangle). So the mission is to find out the rectangle's area, and integrate them.y=0
y=0
and solve the equation to get x = -4 and 4
, so the interval is [-4, 4]
.width
of rectangle is always 2y
.height = 4y
depth
of the box is infinitely thin as dx
.)A(x) = 2y · 4y = 8y² = 8(16 - x²)
[-4, 4]
:
Solve:
x
, so:
Solve:
x = 1 and 5
Disc Method
is a method for calculating the Volume of a 3D shape by rotating a 2D shape
.
The strategy of this method is:
Rotated Circle
, or so called disc
.►Jump to Khan academy for some practice: Disc Method ▼Refer to the article: Finding volumes of 3-D objects with circular symmetry in at least one dimension
Solve:
Disc Method
:
Solve:
y
value, means r = y = eˣ
Area(x) = πr² = π · e²ˣ
Horizontally
along X-axis, so the integral should be ʃ A(x) dx
Volume = ʃ A(x) dx = ʃ π · e²ˣ dx
Solve:
ʃ Area(x) dx
[0, 4]
.r = y - 1 = √x +1 -1 = √x
Area(x) = πr² = πx
[0, 4]
: Volume = ʃ Area(x) dx = ʃ πx dx = 8π
.This method is another kind of Disc Method
, which works on the discs
which its centre is hollow.
Strategy:
Area = A₁ - A₂
:
►Jump to Khan academy for some practice: Washer method ▼Refer to the article: The washer method of calculating volumes of revolution
Solve:
Shell Method
is particularly good for calculating volume of a 3D shape by rotating a 2D shape around a VERTICAL LINE.
Imagine there is a CYLINDER
, and we're to calculate the surface area of the cylinder, and integrate the surface areas when the cylinder gets thiner and thiner.
Refer to Khan academy: Shell method for rotating around vertical line ►Jump to Khan academy for some practice: Shell method ▼Refer to Desmos Animation: Solids of Revolution (about y-axis)
▼Refer to the awesome article: The shell method of finding volumes of revolution
▼Refer to mathdemos.org for more intuitive animations: SHELL METHOD DEMO GALLERY
Solve:
Shell method
of integrating Cylinder's shells
, we got the equation:
►Jump back to previous note: Series (High school level)
Explicit Sequence vs. Recursive Sequence:
Explicit sequence would be presented as:
a𝓃 = a₁ · kⁿ⁻¹
. Recursive sequence would be presented as:a₁ = 3, a𝓃 = k · a𝓃₋₁
Sequence vs. Series:
Sequence is a LIST of numbers, Series is a NUMBER: the SUM of a sequence.
Convergence vs. Divergence:
Convergence means the limit of a function EXISTS. Divergence means the limit DOES NOT EXISTS.
▼Refer to Cool Math: Geometric Series
►Jump to practice: Sequence convergence/divergence
1/3
.►Jump to practice: Finite geometric series
Solve:
Geometric Series formula
, we get the informations as below:
r = -2
n = 20
. Because k
starts from 0, so there're 20 terms.a₀ = -4
Partial sums
is just a fancy word for Finite series
, because it's a a part of infinite series.
Solve:
n
starts from 1, so there're 11 terms, which means we're to calculate S₁₁
.S₁₁ = 88/16 = 11/2
Solve:
a𝓃 = S𝓃 - S𝓃-1
, because S𝓃 = a₁ + a₂ + a₃ +.... + a𝓃-1 + a𝓃
.Evaluate the series, is actually to evaluate the LIMIT of the series function
.
Solve:
Solve:
There are two basic rules for infinite geometric series:
Solve:
|-0.8| < 1
, satisfies the basic converge rules.Solve:
a₀ = 1
r = 0.75
Here lists common Convergence Tests and overview of each. Details are singled out to each section.
Convergence test
are a set of tests to determine wether the series CONVERGENT or DIVERGENT.
It includes:
Test | Description |
---|---|
► Divergent Test | Take nth term's limit. (only to test divergence) |
► Integral Test | Take limit of the series function's integration. |
► p-series Test | Examine at the p value of 1/nᴾ . |
► Comparison Test | Compare the series to a "similar" p-series or geometric-series. |
► Ratio Test | Take limit of two terms ratio. |
► Root Test | Take the limit of nth root of nth term. |
► Alternating Test | Test if terms are decreasing, and take limit of nth term. |
Take the limit of nth term, if it's NOT ZERO, then it's DIVERGENT.
Compare the series to a "similar" p-series or geometric-series.
Compare the series to a "similar" p-series or geometric-series.
Take limit of two terms ratio.
Take the limit of nth root of nth term.
Test if terms are decreasing, and take limit of nth term.
It's also called the
nth term divergence test
. The test can only tell if the series is divergent or not. It CAN NOT tell if it converges.
Jump over to Khan academy practice: nth term test Refer to Khan academy: nth term divergence test
▼Here is the divergent test, very simple:
Solve:
nth term
we get that the limit = 0
.divergent test
rules, we can't conclude anything about it.►Jump over to have practice at Khan academy: Integral test.
Refer to article from tkiryl: The Integral Test
Refer to Khan academy: Integral Test
▼Refer to awesome article from xaktly: Integral Test
Assume the series a𝖓
can be represented as a function f(x)
.
There are a few limitations for it to use the Integral test:
f(x)
MUST BE continuous.f(x) > 0
. It MUST BE a positive function.f'(x) < 0
. It's MUST BE decreasing.Solve:
The Integral test
has introduced the idea of calculating the total area under the function:
Δx = 1
dx
is infinitely small rather than a fixed number Δx = 1
.As been said above, we got this conclusion:
Notice: DO NOT use the
Integral Test
to EVALUATE series, because in general they are NOT equal.
For the the series in form of 1/nᴾ
,
the easiest way to determine its convergence is using the p-series
test:
▼Refer to xaktly: p-series test/harmonic series
Harmonic Series is
∑ 1/n
, and Harmonic Series DIVERGES. That's all you need to know.
You can understand
Comparison Test
intuitively as aSandwich Test
.
THIS TEST IS GOOD FOR RATIONAL EXPRESSIONS
.
Assume that we have a series a_n
, and we're to make up a similar series to it as b_n
:
The logic is:
b > a
& b
converges, then a
converges as well.a > b
& b
diverges, then a
diverges as well.It's so much easier if you think it graphically.
▼Refer to video: Comparison Test (KristaKingMath)
▼Refer to xaktly: Comparison Test
Solve:
a_n
, and we make up a very similar series bigger than it, as b_n
:
p-series test
we get that the b_n
converges.a_n < b_n
, so according to the Sandwich test (Direct comparison test)
, a_n
converges as well.Limit comparison test
is like an extension when the Direct comparison test
won't work.
etc., when we compare a
with b
, although b
converges but a > b
, so we can't make any conclusion.
And that's where the limit comparison test
comes in place.
The logic is:
a/b
.Limit > 0
, then they both converges or both diverges.Limit ≤ 0
, then there's no conclusion.►Jump over to Khan academy for practice: Limit comparison test
▼Refer to video: Limit Comparison Test (KristaKingMath)
▼Refer to xaktly: Limit Comparison Test
Solve:
Solve:
limit comparison test
:
THIS TEST IS GOOD FOR FACTORIALS
.
▼Refer to xaktly: Ratio test / root test
Solve:
ratio test
:
L = 0
, so the series absolutely converges.Solve:
ratio test
, let a_(n+1) / a_n
.(n+1)! = (n+1)·n!
, and (n+8)! = (n+8)·(n+7)!
L = 1
, then there's no conclusion."The root test is used in situations where a series term or part of it is raised to the power of the index variable. "
Notice: The root test isn't a good choice if a series contains factorial terms. If the root test isn't fairly easy to use, you probably shouldn't use it. But when it works, it often cuts to convergence or divergence quickly.
▼Refer to xaktly: Ratio test / root test
Solve:
root test
:
L < 1
, so it converges.It's the test for Alternating series
.
►Refer to Khan academy: Alternating series test ►Refer to xaktly: Alternating Series
It means, Terms of the series "alternate" between positive and negative.
etc., The alternating harmonic series
:
The very good example of this test is the Alternating Harmonic Series
:
▲ It does CONVERGES. (But the Harmonic Series does NOT converge)
Strategy:
Alternating sign (-1)ⁿ
:
Limit = 0
, then the series CONVERGES.Limit ≠ 0
, then the series DIVERGES.Solve:
alternating series
, so we're to apply the alternating series test
.alternating term
, and left with (2/p)ⁿ
.(2/p)ⁿ
is decreasing and its limit is 0
.(2/p) < 1
.p
value, the limit of (2/p)ⁿ
is surely a 0
.p > 2
makes the series converges.Solve:
alternating series
, so we're to apply the alternating series test
.alternating term
, and left with (2n)ᴾ
.(2n)ᴾ
is decreasing and its limit is 0
.p < 0
.p
value, the limit of (2n)ᴾ
is surely a 0
.p < 0
makes the series converges.This section is not about calculation, but rather about the logic.
▶Refer to Khan academy: Conditional & absolute convergence
▶ Back to previous note on: p-series test
▶ Practice at Khan academy: determine-absolute-or-conditional-convergence
▼We can have a series in Given form
and Absolute form
:
We call the series:
Absolute Convergent
: if the series converges in BOTH Given form & Absolute form.Conditional Convergent
: if the series converges ONLY in Given form BUT NOT in Absolute Form.etc.,
Solve:
Solve:
Solve:
Solve:
It's also called the
Remainder Estimation of Alternating Series
.
This is to calculating (approximating) an Infinite Alternating Series:
►Jump over to Khan academy for practice: Alternating series remainder
►Refer to The Organic Chemistry Tutor: Alternate Series Estimation Theorem ►Refer to Mathonline: Error Estimation for Approximating Alternating Series ►Refer to mathwords: Alternating Series Remainder
The logic is:
sum of partial sums & infinite remainder
:
(▲ Sn
is the first n terms, and Rn
is from the n+1 term to the rest terms.)partial sum
& remainder
is:
S & Sn
, so we call it The Error
)The Remainder
MUST NOT be greater than its first term
:
▼Actual sum = Partial sum + Remainder: refer to Khan academy: Alternating series remainder
►Refer to Khan academy: Alternating series remainder ►Refer to Khan academy: Worked example: alternating series remainder
For the Remainder series
, its FIRST TERM
is always DOMINATING the whole remainder:
Based on the error's sign, we could tell the approximated series is UNDERESTIMATED or OVERESTIMATED:
Error > 0
, then the approximated series is Underestimate.Error < 0
, then the approximated series is Overestimate.The error bound
regards to the accuracy of the approximated series, and we want to control the accuracy before approximation.
►Refer to Khan academy: Worked example: alternating series remainder
We have 2 ways to bound the error in a range:
error
to be, orSn
.The Larger n
→ The smaller gap → The lesser Error → The more accurate.
Strategy:
partial sum
to include a certain number of terms, etc. 100 terms
101st term
.error bound
, or the error
is dominated by the first term.error bound
IS the value of the 101st term
.To bound the error in a range, we often say:
What they mean are the same:
▲ And by solving the inequality,
we will get the scope for n
,
then get the Smallest Integer of n
in that scope.
Solve:
partial sum
is already set to 100
terms, so we're to control accuracy by bound the terms
.error
should be from the 101st term
to infinity.error bound
is actually dominated by the first term of the error.error bound = the value of 101st term
:
error bound
is negative, and negative error causes overestimation.Solve:
alternating series
.alternating series test
first, and it passed, which means it converges.partial sum
, and only know how much accurate we'd like.bound the error
, and find out the terms.Error Approximation Theorem
, assume the first term of remainder is a_(n+1)
:
n ≥ 999,999
999,999
the smallest integer of n
to make the series converges with 2 decimal accuracy.The Error Estimation Theorem is not only for alternating series, but available for all infinite series.
▼Boundary of estimating series: refer to Khan academy: Series estimation with integrals
When we see the series as a function, we can actually specify an interval for the function so that the series certainly converges over this interval.
►Jump over to have practice at Khan academy: Interval of convergence
The method is kind of like finding the interval of an ordinary function:
term
of series can't be 0
, so set a_n ≠ 0
and solve it to get the condition.convergence tests
, etc. ratio test
.endpoints
of the interval back in the function and see if it also converges.Solve:
ratio test
which makes it converges:
ratio test
to get the interval:
x
:
endpoints
for this interval:
interval of convergence
is:
Try to think
Power series = Geometric Series
.
►Refer to Math24: Power series
Power series
is actually the Geometric series
in a more general and abstract form.
For easier to remember it, that could be simplified as:
In this function it's critical to know that:
a_n
IS A CONSTANT NUMBER! NOT A VARIABLE !
►Refer to Khan academy: Differentiating power series
We have 2 ways to differentiate series, they work same way:
Either way will do, it depends on the actual equation for you to choose which way you're gonna use.
Solve:
Solve:
x=0
at beginning, everything will be 0
and we don't have anything to calculate.x
in the terms until the last step.expand the series with real numbers
, and we're to try 3 or 4 terms in this case. Third Derivative
, then more than 3 terms will just bring more 0
s.a_n
in the function is:
n
to expand the series:
third derivative
, so let's plug in the x=0
and see what we get:
Solve:
Integrate a series
can be turned to a series of Integrations of terms
:
Geometric series
. So that we can apply the formula of calculating geometric series:
► Jump over to have practice at Khan academy.
Solve:
f(x)
, and the series below is g(x)
g(x) is the
Antiderivativeof the
f(x)`.f(x)
:
antiderivative
can represent the g(x)
, but only with the C
in it:
C
. The easiest way is to plug in 0
for g(x)
:
Solve:
f(x)
and g(x)
.g(x) = -f'(x)
.f(x)
we will get:
-f'(x)
would be:
Taylor series
, or Taylor polynomial
is a series that can REPRESENT a function,
regardless what function it is.
▼Refer to 3Blue1Brown for animation & intuition: Taylor series | Chapter 10, Essence of calculus
"Taylor Series is one of the most powerful tools Math has to offer for approximating functions." - 3Blue1Brown
►Refer to Khan academy: Taylor & Maclaurin polynomials intro (part 1) ▼Refer to xaktly: Taylor Series
(▲ C
represents the centre where we're centred at to approximate the function.)
▲ Notice: The
Taylor Series
is aPower Series
, which means we can use a lot of techniques of power series on this to operate it easily.
We could expand it and make it clearer ▼:
The main purpose of using a Taylor Polynomial
is to REPLACE the original function with a polynomial, which it is easy to work with.
etc., we can express the function f(x) = eˣ
as ▼:
More importantly, by adding more & more terms into the polynomial, we can approximate the function more precisely:
►Refer to joseferrer: Mathematical explanation - Taylor series ►For More animation, visit Desmos: Taylor Series Visualization
Solve:
Taylor Series
centred at x=1
:
(x-1)³
, means all the rest part in the formula, which is:
n=3
, so the coefficient becomes:
f'''(1)
:
Solve:
nth degree
as:
3rd degree
, then it becomes:
It's also called
Maclaurin polynomials
.
Maclaurin series
is a special case of Taylor Series
which centres at x=0
.
▼Expand it we'll understand it better:
▼Here is a graph we're trying to approximate a function centred at x=0
:
Solve:
Maclaurin Series Formula
.x⁴
.x⁴
means we're gonna find out the term of the 4th derivative
, and plug in 4 into the formula, we'll get the term:
x⁴
, which is the rest part in that formula for the term:
4th derivative
.gᴺ(0)
, we can get:
Solve:
Maclaurin series
is a Taylor series centred at x=0
, and the formula is:
f', f'', f'''
and get:
To evaluate a Maclaurin series, we need to convert the series to a function, and then evaluate the function.
▼Jump forward to have a look at the note: Maclaurin Series of Common functions
It's also called the
Lagrange Error Theorem
, orTaylor's Remainder Theorem
.
To approximate a function more precisely, we'd like to express the function as a sum of a Taylor Polynomial & a Remainder.
(▲ For T
is the Taylor polynomial with n terms, and R
is the Remainder after n terms.)
▲Jump back to review the note on Error estimation Theorem.
►Jump over to have practice at Khan academy: Lagrange Error Bound.
The tricky part of that expression is to "preset" the accuracy of the Error
, aka. the Remainder
.
For bounding the Error
, out strategy is to apply the Lagrange Error Bound
theorem.
Simply saying, the theorem is:
(C, x)
:
(▲ for C
is the centre of approximation)z
is any value between C
and x
makes the derivative to the max)
(▲ and note that: the input has to be n+1
)Solve:
Error estimation
method, which first to set up the equation:
error bound
, so we only focus on the Error Rn
.Lagrange Error Bound Theorem
, and bound it to 0.001
:
1.5π
, so C = 1.5π
.1.3π
, so x = 1.3π
.M
value, because all the derivatives of the function cos(x)
, are bounded to 1 even without an interval , so let's say the max value M = 1
.n
except trying some numbers in:
Error
becomes smaller and smaller.n=4
, which means the 4th derivative
, the Error
is less than 0.001
.4th derivative
.Solve:
Lagrange Error Bound Theorem
, and bound it to 0.001
:
0
because it's told as a Maclaurin Series
, so C = 0
.-0.95
, so x = -0.95
.(-0.95, 0)
in this case.M
value, since all the derivatives of eˣ
is just eˣ
, and eˣ
is unbounded at all, so we're to examine the Max value over the interval (-0.95, 0)
Desmos Calculator
, we know that over the interval (-0.95, 0)
, the max value of eˣ
is e⁰ = 1
:
M = 1
.n
to get the desired value:
n=5
and n=6
, we could see that only until n=6
, which means the 6th derivative
, the Error
is less than 0.001
.6th derivative
.Solve:
Lagrange Error Bound Theorem
, and bound it to 0.001
:
2
, so C = 2
.2.5
, so x = 2.5
.(2, 2.5)
.M
value, it's not easy to figure out, but we've been told the formula for derivative.
M
would be:M
expression into the Remainder:
Desmos grapher
we know that when within the interval 2≤ z ≤ 2.5
, that z=2
makes the formula to the max:
n
, we get that n ≥ 3
.►Refer to Khan academy's unit: Finding Taylor or Maclaurin series for a function
Solve:
Infinite Geometric function
:
a = 1
r = -x²
►Refer to Wiki: List of Maclaurin series of some common functions
Maclaurin series of these common functions are very useful, which we really want to memorize.
Function | Maclaurin Series |
---|---|
sin(x) | |
cos(x) | |
tan(x) | |
sec(x) | |
eˣ | |
Geometric series 1 | |
Geometric series 2 | |
Geometric series 3 |
Solve:
cos(x) Maclaurin series
:
Solve:
Solve:
factorial dominators
, we found that resembles the pattern of the sin(x)
's maclaurin series.sin(x) maclaurin series
, we notice that x = π/2
, and all the rest are the same.1
.Solve:
"The
Euler's identity
connects all of these FUNDAMENTAL NUMBERS in some mystical way that shows that there's some connectedness to the UNIVERSE. If this does not blow your mind, you have no emotion." - Sal Khan
▶︎Refer to the most well-known lecture from Sal Khan: Euler's formula & Euler's identity
▼ Euler's formula
:
▼ Euler's identity
:
"...this concept is nothing to be intimidated by. Partial differentiation is essentially just taking a multi dimensional problem and pretending that it's just a standard 1D problem when we consider each variable separately."
Now we take many partial derivatives
out from a function, and we need to pack them together in a form.
We call the form of pack The gradient of the function
.
For example:
It is the more general form of the
Partial derivative
, because it not only give the derivatives with respects to axes, but to any direction.
Refer to Khan academy: Directional derivative Refer to youtube: Directional Derivatives and The Gradient
The Partial derivatives
only give the SLOPE in x direction
and y direction
and more directions with respect to the variables, but what if we want to have a SLOPE in ANY DIRECTION?
So the Directional derivative
comes in place to achieve that.
How to do this:
θ
unit vector
in form of u = cosθ·i + sinθ·j
, because cos²θ + sin²θ = 1
.Dᵤ
means Directional Derivative of a Unit Vector u
, f𝓍
means Partial derivative respects to x, and f𝑦
means Partial derivative
respects to y.)The Directional derivative
is the Dot product
of The Gradient & Unit vector
:
Which comes from:
Study resources
Study Tools
Practice To-do List (Linked with Unit tests)
Table of Contents
Basic Differential Rules
Basic Integral Rules