Mathemation · YouTube · 9:48
This 150-second whiteboard lecture introduces polynomial interpolation as the problem of finding a simple curve through a finite set of data points when the exact underlying function is unknown. The speaker first explains why exact recovery between points is impossible without assumptions, then narrows the search from an arbitrary smooth curve to a polynomial. A visual sequence on a scatter plot motivates the rule that n points call for a polynomial of degree n-1. The final part switches to a concrete example: fitting a cubic +bx+cx^ through four points with x-values 0,1,2,3, and deriving the corresponding 4-equation linear system for a,b,c,d. The clip stops before solving the system. This 150-second whiteboard segment sets up the problem of fitting a cubic polynomial + bx + cx^ through four points with x-values 0, 1, 2, 3. Substituting the points gives a 4-equation linear system in a, b, c, d. The narrator then presents Newton’s finite-difference approach: subtract successive equations to eliminate the common a term, producing Delta , Delta , and Delta . This reduces the problem to a 3-equation system in b, c, d. The clip ends by noting that the same idea can be applied again to eliminate b, but the second-difference step itself is not carried out within this segment. This video segment demonstrates the method of finite differences for solving a system of linear equations derived from fitting a cubic polynomial to four data points. By repeatedly taking differences of the equations, the presenter systematically eliminates the coefficients 'b' and 'c', eventually isolating 'd'. The segment concludes by showing how to express all coefficients (a, b, c, d) in terms of the finite differences of the y-values. This clip derives the Newton forward interpolation formula first for four equally spaced points and then generalizes it to points. It begins from a cubic polynomial +bx+cx^, rewrites a,b,c,d using forward differences, and obtains the boxed formula . The second half states the analogous n-th order formula for points (0,),...,(n,), explains the pattern of falling-factor products, increasing difference order, and factorial denominators, and warns that this version applies only when the x-values are exactly 0,1,...,n and that the last product stops at x-(n-1).
Use the learning inspector for key ideas and moments, or open the reading tabs for the complete notes.
Generated from the video's visuals and explanation; not verbatim speech.
The clip opens on a blackboard-style slide titled "Polynomial Interpolation - Overview," with a reference box reading "Notes: p. 89-94" and "Rogawski: ." On the left, the problem is written: "Given a set of data points, can we find a smooth curve passing through them?" On the right, four yellow points are plotted in the plane. The speaker introduces the topic as a weekly overview and tells viewers to consult the course notes because the material is not in the e-text.
The narration then frames the mathematical situation: suppose there is an unknown function , but its exact formula is unavailable. Instead, only a few measured data points are known. The speaker gives applied examples such as speed over time or daily profit, emphasizing that the same abstract question arises in many contexts.
From those limited samples, the goal is to estimate what the function is doing at intermediate inputs. The speaker immediately warns that this cannot be answered with certainty: between the plotted points, the true function could behave in many different ways. This establishes the need for modeling assumptions rather than exact recovery.
Under the assumption that the function is reasonably well behaved and that the sampled points reflect its general trend, the video proposes approximating the missing behavior by finding a curve through the data. At this stage the board still says "smooth curve," matching the spoken formulation.
The method is then simplified. The speaker says they will not use just any smooth curve; instead, they will choose the simplest useful family, namely polynomials. Visually, the word "smooth" is crossed out and replaced by "polynomial," turning the problem into: find a polynomial curve passing through the given points.
A sequence of sketches on the graph motivates the degree choice. For two points, a line suffices. For three points, the speaker suggests a parabola, i.e. a polynomial of degree 2. For four points, a cubic, i.e. degree 3, is indicated. The board then summarizes the pattern as "polynomial of degree n-1" for points.
The scene changes to a new board titled "Example: Fitting a Cubic Through 4 Points." The goal is to find a cubic polynomial passing through , , , and . Substituting those four -values into the cubic yields the displayed system: , , , and . The clip ends here, before the system is solved.
The segment opens on a black digital whiteboard with the title “Example: Fitting a Cubic Through 4 Points.” The stated goal is to find a cubic polynomial that passes through , , , and . The unknowns are the four coefficients .
The narrator explains the basic constraint: the cubic must satisfy the polynomial equation at each given point. Substituting into makes the , , and terms vanish, leaving . Repeating the same substitution for gives the displayed system , , and .
At this stage the problem has been converted into a linear system in . The narrator notes that modern linear algebra could solve it, but introduces Newton’s historical alternative: exploit the fact that every row contains exactly one term.
Newton’s method proceeds by subtracting successive equations. Define . Using the first two rows, , so . Similarly, , and .
These three new equations form a reduced system with three unknowns, , and no remaining . The narrator names as first finite differences. He then observes that each reduced equation contains a single term, so the same subtraction strategy can be applied again to eliminate next. The clip ends before writing the second-difference system.
The video starts by presenting a system of three equations representing the first finite differences of a cubic polynomial: , , and . The goal is to eliminate the variable .
To eliminate , the presenter takes the difference between consecutive equations. Subtracting the first equation from the second gives the second finite difference: .
Similarly, subtracting the second equation from the third yields another second finite difference: . The terms have now been successfully eliminated.
Next, the presenter eliminates the term by taking the difference of the second finite differences. This results in the third finite difference: .
With only remaining, it can be easily solved: . Since the values are known data points, their finite differences can be computed, allowing to be determined.
Once is known, the presenter explains that one can work backward through the equations to find , then , and finally . The explicit formulas for and in terms of the finite differences are displayed for completeness.
The clip opens on a blackboard already containing the finite-difference setup for four data points. On the right is the cubic template , and on the left are relations expressing in terms of .
The presenter explains that once are replaced by the derived expressions, the polynomial is no longer written with unknown coefficients but directly in terms of the given -values.
The solved coefficient formulas are visible: , , , and .
Substituting these into the cubic and collecting terms yields the boxed four-point Newton forward interpolation formula .
The narration emphasizes the meaning of each piece: the constant term is , the next term uses the first forward difference, then the second, then the third, with denominators 2 and 6.
The result is presented as a general fact for any four data points located at : this cubic expression passes through all four points.
The presenter then invites the viewer to inspect the structure of the formula for patterns that could extend beyond four points.
The board changes to a new heading, Polynomial Interpolation: Points, and states that the -th order polynomial through is given by the displayed series.
The generalized formula is .
The explanation of the pattern is explicit: each new term adds another descending -factor, advances to the next forward difference, and divides by the matching factorial.
Before using the formula in examples, the presenter gives a restriction: this version applies only when the data points have -values exactly .
A second remark clarifies a common structural confusion: although the data run up to , the final product in the formula stops at , not .
The reason given is that the role of the endpoint is already absorbed into computing the last forward difference , so no extra factor appears in the expression itself.
The video poses the basic question: given a finite set of data points, can we find a curve passing through them so as to approximate an unknown function between the samples? This is presented as the motivating problem before any formula is introduced.
The speaker states that from only a few data points one cannot know with certainty what the underlying function does in between. Interpolation therefore requires assumptions about regularity and trend rather than claiming exact reconstruction.
To keep the model simple, the video narrows the search from an arbitrary smooth curve to a polynomial. This is shown directly on the board by crossing out "smooth" and writing "polynomial."
Using the examples of 2, 3, and 4 points, the video states the pattern that points require a polynomial of degree . The board writes this as "polynomial of degree n-1."
The worked example asks for a cubic polynomial through four points with x-values 0, 1, 2, and 3. The candidate form is , introducing four unknown coefficients to be determined.
Imposing passage through , , , and gives a four-equation linear system in . This is the algebraic setup for solving the cubic interpolant.
The example seeks a cubic polynomial that passes through four points whose x-values are . The coefficients are the unknowns.
Imposing the four interpolation conditions gives one equation per point. Substitution yields , , , and .
Because each original equation contains exactly one term, subtracting successive equations cancels . This produces the reduced system , , , a -equation system in .
The narrator points out that the reduced system still has a single in every equation, so the same difference trick can be repeated to eliminate . The actual second-difference calculation is not shown in this clip.
The initial system of equations derived from the cubic polynomial, representing the first finite differences.
Calculated by subtracting the first equation from the second, eliminating the term.
Calculated by subtracting the second equation from the third, also eliminating the term.
Calculated by subtracting the two second finite differences, eliminating the term to isolate .
The coefficient is found by dividing the third finite difference by 6.
The explicit formulas for and expressed in terms of the finite differences of the values.
The video begins with a general cubic polynomial and treats its coefficients as unknowns to be rewritten using finite differences. This is the starting point for deriving the four-point interpolant.
By solving the displayed difference relations, the coefficients of the cubic are rewritten entirely in terms of and the forward differences at index 0.
After substitution, the interpolating cubic takes a structured form built from falling-factor products and successive forward differences. The presenter states that this polynomial passes through any four data points at .
The same pattern extends from four points to equally spaced points. Each term increases the number of descending factors, raises the order of the forward difference, and divides by the corresponding factorial.
The presenter explicitly restricts this version of the formula to data whose -values are exactly . Other node placements require a later generalization.
Even though the data include the point with , the final falling-factor product in the formula ends at . The endpoint contributes through the computation of , not through an extra factor .
Explore conditions, steps and evidence. Supplementary explanations are labeled separately from content shown in the video.
The speaker says, "Suppose that you're working with some function y equals f of x".
An underlying function whose exact equation is not known.
Real-valued function context; the video does not state a precise domain or range.
The speaker describes having "a few data points" collected through experimentation.
Four yellow plotted points are shown on a coordinate graph.
n
Number of given data points in the general interpolation discussion.
Positive integer; the video uses n in the phrase "with n points" without further formal definition.
The speaker says, "So in general, with n points, we're going to need a polynomial of degree n minus one."
The board writes "polynomial of degree n-1".
n-1
Degree of the polynomial proposed for interpolating n points.
Integer degree associated with n data points.
The board shows " + bx + cx^".
+bx+cx^
General cubic polynomial used as the interpolation target in the example.
Polynomial in x with coefficients a,b,c,d; the video does not specify coefficient field beyond ordinary algebraic notation.
The coefficients appear in the displayed cubic formula and in the four linear equations.
a,b,c,d
Unknown coefficients of the cubic polynomial to be solved from the data points.
Scalar coefficients in the displayed system.
The board lists "(0,), (1,), (2,), (3,)".
The speaker names the same four points.
(0,),(1,),(2,),(3,)
Four interpolation nodes with equally spaced x-values 0,1,2,3 and unknown y-values ,,,.
x-values fixed at 0,1,2,3; treated as given data values.
The board displays the four equations , , , .
Linear system obtained by substituting ,1,2,3 into the cubic polynomial.
Equations relating the unknown coefficients a,b,c,d to the data values ,,,.
The board shows the cubic polynomial as + bx + cx^.
The narrator says, "y equals a plus b x plus c x squared plus d x cubed."
y
Dependent variable or output value of the interpolating cubic polynomial.
Real-valued function output.
The polynomial is written in powers of x: + bx + cx^.
The narrator plugs in for the first point and refers to x-values 0, 1, 2, 3.
x
Independent variable of the cubic polynomial; the interpolation nodes are , 1, 2, 3.
Real-valued input variable.
a appears as the constant coefficient in + bx + cx^ and in the system , , etc.
The narrator asks how to determine the coefficients a, b, c, d and later says each row has a single a term.
a
Constant coefficient of the cubic polynomial; after substitution it equals .
Unknown scalar coefficient.
b appears as the coefficient of x in + bx + cx^ and in the derived first-difference equations.
The narrator lists b among the unknown coefficients and notes that each new equation contains a b term.
b
Coefficient of the linear term x in the cubic polynomial.
Unknown scalar coefficient.
c appears as the coefficient of in the polynomial and with multipliers 1, 3, 5 in the first-difference system.
The narrator lists c among the unknown coefficients.
c
Coefficient of the quadratic term in the cubic polynomial.
Unknown scalar coefficient.
Title reads "Polynomial Interpolation - Overview".
Problem statement on board: "Given a set of data points, can we find a smooth curve passing through them?"
Speaker explains wanting to estimate what the function is doing at points in between using limited information.
The video defines the central problem as: given a set of data points, find a smooth curve passing through them so that one can approximate the behavior of an unknown function between the sampled points.
Data points are known.
The exact equation of the underlying function is not known.
The goal is approximation between the given points.
Speaker says they will not use just any old smooth curve but will keep it as simple as possible and use a polynomial.
The word "smooth" is crossed out and replaced with "polynomial"; later the board adds "of degree n-1".
To make the interpolation problem tractable, the video restricts the desired smooth curve to a polynomial, specifically choosing the simplest polynomial family that can pass through the given points.
One wants a simple approximating curve.
The curve must pass through the given data points.
Speaker gives examples: two points can use a line, three points may use a parabola, four points likely need a cubic, and in general n points need degree n-1.
Board text becomes "polynomial of degree n-1".
The video states a pattern-based rule: with n points, one should use a polynomial of degree n-1. It illustrates this with 2 points → line, 3 points → parabola, 4 points → cubic.
There are n data points.
The intended interpolant is a polynomial chosen by the degree pattern shown in the video.
New board title: "Example: Fitting a Cubic Through 4 Points".
Goal line: "Find a cubic polynomial + bx + cx^ Passing through (0,), (1,), (2,), (3,)."
The example begins by specifying a cubic polynomial with four unknown coefficients and requiring it to pass through four data points whose x-coordinates are 0, 1, 2, and 3.
There are exactly four data points.
The x-values are 0,1,2,3.
The desired interpolant is cubic.
Title: Example: Fitting a Cubic Through 4 Points. Goal: Find a cubic polynomial + bx + cx^ passing through (0,), (1,), (2,), (3,).
The narrator states the goal and explains that substituting the four points yields a system of four equations with four unknowns.
The clip defines the task of finding a cubic polynomial that passes through four specified points with equally spaced x-values 0, 1, 2, 3. The unknowns are the coefficients a, b, c, d, and the point conditions produce a linear system in those coefficients.
The polynomial is cubic.
The four given points have x-coordinates 0, 1, 2, 3.
The coefficients a, b, c, d are unknowns to be determined.
The board displays the system: ; ; ; .
The narrator explains that plugging in each point gives one equation, ending with a system of four equations and four unknowns.
Each interpolation condition is substituted into the cubic form. Because the x-values are 0, 1, 2, 3, the powers of x become , , , , producing the displayed coefficient multipliers for b, c, and d.
Use the same cubic form + bx + cx^.
Substitute , 1, 2, 3 respectively.
Treat , , , as known data values.
The board adds Delta , Delta , Delta .
The narrator says Newton observed each row has a single a term, so taking differences of successive equations kills all the a terms; he calls these first finite differences.
Newton’s method in this clip subtracts each equation from the next one. Since every original equation contains exactly one a term, the subtraction removes a and leaves a smaller linear system in b, c, d. The resulting quantities Delta , Delta , Delta are named first finite differences.
Start from the four-equation system obtained from the cubic interpolation constraints.
Subtract successive equations: equation 2 minus equation 1, equation 3 minus equation 2, equation 4 minus equation 3.
The original system must have a single a term in each row.
The speaker explains taking finite differences to eliminate variables in a system of equations.
Equations for , , , , , and are written out.
The method of finite differences involves calculating the differences between consecutive values of a sequence. By repeatedly taking these differences, one can eliminate variables in a system of polynomial equations, eventually isolating the highest-degree coefficient.
Applicable to polynomial sequences.
The general form of a cubic polynomial + bx + cx^ is shown.
A cubic polynomial is a polynomial of degree three. It has the general form + bx + cx^, where a, b, c, and d are constants, and d is not equal to zero.
The right side of the board shows + bx + cx^.
The speaker says replacing a,b,c,d with the derived expressions gives a polynomial fitting four data points.
For four data points, the video starts from a general cubic polynomial in x and then substitutes finite-difference expressions for its coefficients so that the resulting polynomial is expressed directly in terms of the given y-values.
Applies to the four-point case discussed in this clip.
The data points are taken at ,1,2,3.
The board displays , , , then , , and .
Below them are , , , .
The cubic coefficients are rewritten using first, second, and third forward differences at index 0. This converts the unknowns a,b,c,d into quantities computable from the data values themselves.
Four consecutive data values are available.
The x-values are 0,1,2,3.
The bottom boxed formula is .
The speaker reads the formula term by term and says it passes through all four points for x-values 0,1,2,3.
After substituting the coefficient expressions into the cubic template, the interpolating polynomial is written as a sum of falling-factorial terms multiplied by successive forward differences divided by factorials.
Exactly four data points are used.
Their x-values are 0,1,2,3.
Speaker says the problem is impossible to answer with certainty because the function could be doing all sorts of things in between the points.
From only a finite set of data points, one cannot determine with certainty what the underlying function does between those points.
The exact equation of the function is unknown.
Only a finite set of data points is available.
For a function known only at finitely many sample points.
Speaker says, "So in general, with n points, we're going to need a polynomial of degree n minus one."
Board writes "polynomial of degree n-1".
The video presents this as a general rule but does not prove uniqueness or discuss exceptional configurations such as collinear points.
With n points, the video states that one needs a polynomial of degree n-1 to interpolate them.
There are n data points.
The interpolant is chosen according to the pattern illustrated in the video.
Stated generally for n points, without additional hypotheses in the clip.
The narrator says that if you know a little linear algebra you should be able to solve the system, but when Sir Isaac Newton considered this problem, linear algebra had not yet been invented, so he solved it in a slightly different way.
The clip presents Newton’s finite-difference approach as an alternative to solving the 4-by-4 linear system by modern linear algebra, motivated historically by the fact that linear algebra had not yet been invented in Newton’s time.
The problem is the cubic-through-four-points system shown on the board.
The comparison is between modern linear-algebra solution methods and Newton’s difference method.
The statement is presented as a historical explanation for the method used in this example.
After the three Delta equations are written, the board shows a new grouped system with three equations in b, c, d.
The narrator says, "By considering these delta y terms, which by the way are called first finite differences of our system, we end up with a new system with three equations and three unknowns."
Taking first finite differences of the original four-equation system produces a new system with three equations and three unknowns, namely b, c, and d.
The original system is the four-equation cubic interpolation system.
Successive equations are subtracted as shown.
For the displayed system, applying the first-difference operation yields exactly three reduced equations.
The narrator says there is a b in each equation in the new system, which means he could probably do the exact same thing to the new system, take differences of successive equations, and kill off yet another variable.
The reduced equations visibly all contain a b term: , , .
The clip announces the next step but does not actually perform the second-difference calculation within this segment.
Because each equation in the first-difference system contains a single b term, the same subtraction strategy can be applied again to eliminate b and reduce the system further.
The reduced system is the three-equation system in b, c, d shown after the first differences.
Each row contains exactly one b term.
The claim is stated prospectively for the displayed reduced system; the actual second-difference equations are not shown in this clip.
The speaker says that given any four data points for x-values 0,1,2,3, we can find a cubic polynomial given by the boxed expression that passes through all four.
For any four data points with x-values 0,1,2,3, the displayed cubic expression gives a polynomial that passes through all four points.
Four data points are given.
Their x-values are 0,1,2,3.
Universal over arbitrary data values ,,, at those x-positions.
The speaker says the exact same methods used for four points can be used to find a polynomial of order n passing through data points.
The board writes the n-th order polynomial through (0,),...,(n,).
The same finite-difference construction extends from four equally spaced points to equally spaced points, yielding an n-th order interpolating polynomial.
There are data points.
Their x-values are 0,1,...,n.
For positive integer n and arbitrary data values ,...,.
The speaker explicitly says this formula only applies when your data points have x-values 0,1 all the way up to n.
The displayed generalized formula applies only when the data points have x-values exactly 0,1,...,n.
The formula being referenced is the one shown on the board for points.
Conditional statement about use of the formula.
The speaker says we never actually multiply by x-n; we stop at x-(n-1).
The last displayed product ends with (x-(n-1)) before !.
In the generalized formula, although the data run up to , the final product factor is x-(n-1), not x-n.
Using the generalized -point formula shown on the board.
Statement about the structure of the last term.
The board shows the cubic +bx+cx^ and then the four equations obtained from the points (0,),(1,),(2,),(3,).
Start with the general cubic polynomial stated as the goal of the example.
Displayed directly on the board as the target form.
Substitute into the cubic and set the result equal to the first data value .
Point (0,) lies on the polynomial; powers of 0 eliminate b,c,d terms.
Substitute into the cubic and equate to .
Point (1,) lies on the polynomial; and .
Substitute into the cubic and equate to .
Point (2,) lies on the polynomial; and .
Substitute into the cubic and equate to .
Point (3,) lies on the polynomial; and .
Requiring the cubic to pass through the four given points produces the displayed 4-by-4 linear system in the unknown coefficients a,b,c,d.
Speaker lists 2 points → line, 3 points → parabola, 4 points → cubic, then generalizes to n points → degree n-1.
The graph successively shows a line through two points, a parabola through three points, and a cubic-like curve through four points.
This is an illustrative pattern rather than a proof in the clip.
The speaker notes that two points can be joined by a line.
Directly stated in the audio and shown visually.
For three points, the speaker suggests a parabola, i.e. a polynomial of degree 2.
Directly stated in the audio and shown visually.
For four points, the speaker says one would likely need a cubic, i.e. degree 3.
Directly stated in the audio and shown visually.
The speaker generalizes the observed pattern to n points requiring degree n-1.
Explicit verbal generalization and matching board text.
The video motivates the rule "n points require a polynomial of degree n-1" by extending the visible pattern from 2, 3, and 4 points.
The narrator says, "if I take any of the four points, let's say the first one, and I plug in zero for my x, then I should get y naught on the left hand side. But if I plug in zero for x, my last three terms disappear, and I just get y naught is equal to a."
The first equation on the board is .
Start from the general cubic polynomial whose coefficients are to be found.
This is the stated interpolation model on the board.
Use the first interpolation point (0, ).
The problem requires the cubic to pass through (0, ).
Substitute and into the cubic.
Direct substitution into the polynomial equation.
All terms containing x vanish because they are multiplied by 0.
Arithmetic simplification using 0 powers and multiplication by 0.
The first interpolation condition gives .
The board displays , , alongside .
The narrator says he does the exact same thing with the other terms and ends up with a system of four equations with four unknowns.
The clip does not verbally expand each substitution for , 2, 3 step by step; it summarizes the pattern after demonstrating .
The cubic must satisfy the polynomial equation at each of the four given points.
These are the interpolation constraints stated in the goal.
Substitute and into the cubic.
Direct substitution into + bx + cx^.
Simplify the powers of 1.
Since for ,2,3.
Substitute and into the cubic.
Direct substitution into the polynomial model.
Evaluate the powers 2, , and .
Arithmetic simplification: , , .
Substitute and into the cubic.
Direct substitution into the polynomial model.
Evaluate the powers 3, , and .
Arithmetic simplification: , , .
The four point conditions produce the displayed 4-by-4 linear system in a, b, c, d.
The board successively writes Delta , Delta , Delta .
The narrator explains that subtracting one equation from the previous one kills the a term in each case.
Define the first difference using the first two original equations.
Stated definition in the clip.
Substitute the expressions for and from the original system.
Use and .
Cancel the a terms.
Algebraic subtraction; each original row contains a single a term.
Define the next first difference using equations two and three.
Stated definition in the clip.
Substitute the original expressions for and .
Use the displayed system rows for and .
Collect like terms after cancellation of a.
, , .
Define the last first difference using equations three and four.
Stated definition in the clip.
Substitute the original expressions for and .
Use the displayed system rows for and .
Collect like terms after cancellation of a.
, , .
The first finite differences yield the reduced system , , .
The speaker explains subtracting the first equation from the second to eliminate the 'b' term.
The derivation shows .
First finite difference equation.
Given
Second finite difference equation.
Given
Definition of second finite difference.
Definition
Substituting the expressions for and .
Substitution
Simplifying the expression eliminates the 'b' term.
Algebraic simplification
Taking the second finite difference eliminates the 'b' term, leaving an equation with only 'c' and 'd'.
The speaker explains subtracting the second finite difference equations to eliminate the 'c' term.
The derivation shows .
First second finite difference equation.
Derived previously
Second second finite difference equation.
Derived previously
Definition of third finite difference.
Definition
Substituting the expressions for and .
Substitution
Simplifying the expression eliminates the 'c' term.
Algebraic simplification
Taking the third finite difference eliminates the 'c' term, leaving an equation with only 'd'.
Top-left system expresses ,, in terms of b,c,d; middle-right expresses , in terms of c,d; far-right gives .
Lower-left solves for d,c,b,a; lower-right rewrites y as the boxed Newton form.
The speaker says to replace a,b,c,d with the expressions just derived to obtain the polynomial in terms of the y-values.
The intermediate algebraic simplification from the substituted coefficients to the compact falling-factorial form is not fully written out step by step on screen.
Start with a general cubic polynomial for the four-point case.
Displayed on the board and spoken by the presenter.
Write the first forward differences in terms of the unknown coefficients.
Shown explicitly in the yellow system at top left.
Form second differences by subtracting adjacent first differences.
Shown explicitly in the purple system.
Form the third difference from the second differences.
Shown explicitly at the upper right.
Solve backward from the highest difference to express every coefficient using forward differences and .
Written directly on the board in cyan.
Substitute those coefficient expressions into the cubic and collect terms into the Newton forward-difference form.
The final boxed formula is shown, and the audio states that replacing a,b,c,d yields the interpolating polynomial.
The four-point cubic interpolant can be written directly in terms of and the first three forward differences at index 0.
The speaker says patterns in the equation suggest generalizing to n points, not just four.
The next board replaces the specific 2 and 6 denominators with 2!,3!,...,n! and extends the product to x-(n-1).
This is presented as a pattern-based generalization rather than a full proof of the general formula within this clip.
Inspect the four-point formula for repeating structure.
Boxed formula remains visible before the board changes.
Each new term multiplies by one additional descending factor.
Explicitly described in the audio and visible in the formula.
Each new term uses the next higher forward difference.
Explicitly described in the audio and visible in the generalized formula.
The denominators follow factorial order matching the difference order.
The speaker says we divide by k factorial up to n factorial; the board shows 2!,3!,...,n!.
Continue the observed pattern to obtain the formula for points.
Written on the second board after the transition.
The four-point formula suggests a general -point Newton forward interpolation formula by extending the same structural pattern.
Board title: "Example: Fitting a Cubic Through 4 Points".
Goal and system are fully displayed on the board.
Speaker introduces the example with four data points and says they want to find a cubic polynomial.
The clip ends before solving the system or giving numerical y-values.
Find a cubic polynomial that passes through the four data points (0,), (1,), (2,), and (3,).
The interpolant is assumed to have the form +bx+cx^.
The x-coordinates of the data points are 0,1,2,3.
The corresponding y-values are denoted ,,,.
Set up the equations that determine the coefficients a,b,c,d of the cubic interpolant.
Write the general cubic polynomial as the candidate interpolant.
Stated explicitly in the example goal on the board.
Substitute ,1,2,3 into the cubic and equate each result to the corresponding y-value.
Each listed point must lie on the polynomial; substitution yields the displayed system.
The example is set up as the linear system .
No numerical verification is performed within the clip; the displayed equations are consistent with direct substitution of ,1,2,3 into +bx+cx^.
The entire board is organized around the example titled "Example: Fitting a Cubic Through 4 Points" with the polynomial, the four points, the original system, and the first-difference system.
The narrator walks through setting up the problem, deriving the system, and applying Newton’s first-difference reduction.
No numerical values for , , , are supplied, so the coefficients are not computed explicitly in this clip.
The clip stops before performing the announced second-difference step.
Find a cubic polynomial + bx + cx^ passing through (0,), (1,), (2,), (3,).
Four interpolation points with x-coordinates 0, 1, 2, 3.
Corresponding y-values denoted , , , .
Unknown coefficients a, b, c, d.
Determine the coefficients of the cubic by converting the interpolation conditions into a solvable linear system and then reducing that system using finite differences.
Write the general cubic interpolant.
This is the stated model for the example.
Substitute the four points into the cubic to obtain four linear equations in the four unknown coefficients.
Each point must satisfy the polynomial equation.
Subtract the first equation from the second to eliminate a.
Each original row contains exactly one a term, so successive subtraction cancels a.
Subtract the second equation from the third to eliminate a again.
Same finite-difference elimination principle applied to the next pair of rows.
Subtract the third equation from the fourth to eliminate a once more.
Same finite-difference elimination principle applied to the last pair of rows.
Collect the results into a smaller system with three equations and three unknowns.
The narrator explicitly identifies this as the reduced system after first finite differences.
The clip reduces the original 4-by-4 interpolation system to the 3-by-3 first-difference system in b, c, d. It does not finish solving for the coefficients within this segment.
The reduction can be checked by direct subtraction of the displayed original equations; each resulting right-hand side matches the board entries , , and .
Left side shows the title and problem statement; right side shows a coordinate plane with four yellow points.
A red pointer moves among the title, notes reference, problem text, and plotted points.
Title "Polynomial Interpolation - Overview"
Reference box "Notes: p. 89-94" and "Rogawski: "
Problem statement text
Coordinate axes labeled x and y
Four yellow data points
Red pointer dot
The pointer highlights the title and notes reference early in the segment.
The pointer moves over the plotted points while the speaker discusses unknown functions and intermediate behavior.
The word "smooth" is crossed out and replaced by "polynomial".
The phrase is extended to "polynomial of degree n-1".
The four plotted points remain in the same positions throughout the overview section.
The overall left-text/right-graph layout remains unchanged until the scene switch.
The visual layout frames interpolation as a geometric problem: given scattered points in the plane, seek a simple curve passing through them.
The graph successively shows a straight line through two points, then a parabola-like curve through three points, then a cubic-like curve through four points.
The narration matches these cases: line for two points, parabola for three, cubic for four.
Exact drawing times are approximate because the curves appear during continuous narration rather than as separately timestamped labels.
Two-point line sketch
Three-point parabola sketch
Four-point cubic-like sketch
Original four yellow data points
A line is drawn through the first two points.
It is replaced by a parabola passing through three points.
It is then replaced by a higher-degree curve passing through all four points.
The underlying data points do not move.
The coordinate axes remain fixed.
The animation visually motivates increasing polynomial degree as the number of interpolated points increases.
At 131 seconds the board changes to a new example page titled "Example: Fitting a Cubic Through 4 Points".
The new page contains the cubic formula and a four-equation system.
New title "Example: Fitting a Cubic Through 4 Points"
Goal statement with cubic formula
List of four points
Large brace grouping four equations
The previous graph-and-problem layout disappears.
A text-only example layout replaces it.
The red pointer moves across the listed points and equations near the end of the clip.
The topic remains polynomial interpolation.
The example continues the prior claim that four points lead to a cubic.
The scene change marks the transition from conceptual overview to algebraic setup of a concrete interpolation example.
A black digital whiteboard shows the title at top, the goal statement in green, the four points in green, and a yellow brace grouping four yellow equations on the left.
A red pointer dot moves among the polynomial terms, the listed points, and the equations while the narrator explains substitution.
Title text
Goal statement
Cubic polynomial + bx + cx^
Four points (0,), (1,), (2,), (3,)
Yellow system of four equations
Red pointer dot
The pointer highlights the polynomial coefficients and then the point list.
Attention shifts from the general cubic to the specific equations generated by substitution.
The original four-equation system remains visible on the left throughout this interval.
The x-values of the points remain fixed at 0, 1, 2, 3.
The visual organization separates the problem statement from the algebraic consequences of imposing the interpolation conditions.
Purple braces and purple equations are added to the right of the yellow system one by one.
The new lines read Delta , then Delta , then Delta .
Original yellow system
Purple braces
Purple Delta y equations
Red pointer dot
A purple brace is drawn beside the first two yellow equations and the first Delta equation is written.
A second purple brace and Delta line are added below.
A third purple brace and Delta line are added below those.
The original yellow system stays unchanged on the left.
Each new purple equation corresponds to subtraction of successive original equations.
The color separation visually distinguishes the original constraints from the derived reduced system produced by Newton’s finite-difference method.
By the end of the clip, the right side contains three stacked purple equations grouped as a new system.
The narrator points out that the new system has three equations and three unknowns and that each equation contains a b term.
The clip does not draw the second-difference system before ending.
Three purple first-difference equations
Remaining yellow original system
Red pointer dot
The narrator’s attention shifts from the act of subtraction to the structure of the reduced system.
The possibility of repeating the method is introduced verbally.
The reduced system remains in terms of b, c, d only.
No numerical y-values are introduced.
The final board state demonstrates the key payoff of the method: a smaller linear system that is structurally similar enough to suggest iterative elimination.
Purple brackets are drawn to group equations together for subtraction.
Equations
Brackets
Brackets are drawn around pairs of equations to indicate which ones are being subtracted.
The equations themselves remain unchanged.
The visual grouping helps the viewer follow the step-by-step elimination process.
At about 66 seconds the entire blackboard content changes from the four-point derivation to a new title and generalized formula.
Previous four-point derivation
New title Polynomial Interpolation: Points
Generalized formula box
All prior equations disappear at once.
A new heading and a single generalized formula replace them.
The topic remains polynomial interpolation.
Forward-difference notation continues to be used.
The visual transition marks the move from a concrete four-point example to the abstract -point formula.
A small red dot moves among parts of the generalized formula and the listed points while the presenter discusses restrictions and the final factor.
Exact frame-by-frame path of the dot is approximate because it moves continuously.
Point list (0,),(1,),...,(n,)
Last term ...(x-(n-1))!
The dot shifts attention from the data-point list to the final product and final difference term.
The formula itself does not change during this emphasis sequence.
The animation supports the spoken remarks that the formula requires x-values 0 through n and that the product stops at x-(n-1).
Speaker warns that the problem is impossible to answer with certainty because the function could behave arbitrarily between points.
One might think that a few data points determine the original function uniquely everywhere between them.
The video explicitly says this is impossible with certainty; interpolation only provides an approximation under additional assumptions about well-behavedness and trend.
Speaker says they are not going to use just any old smooth curve and will keep it as simple as possible.
The board crosses out "smooth" and writes "polynomial".
One might assume the problem statement permits any smooth curve whatsoever as the interpolant.
The video narrows the choice to polynomials in order to keep the model simple and computationally manageable.
The narrator says that with a little linear algebra one should be able to solve the system, but Newton solved it differently because linear algebra had not yet been invented.
One might think this interpolation problem must be solved using modern linear-algebra machinery such as matrices and elimination theory.
The clip shows an alternative historical route: repeatedly subtract successive equations to eliminate shared leading terms, producing smaller systems without invoking formal linear algebra.
The narrator emphasizes that taking differences kills the a terms and creates a new system with three equations and three unknowns.
The board transforms a 4-equation system into a 3-equation system via Delta y definitions.
One might think Delta , Delta , Delta are just relabelings that do not change the problem.
In this clip they are an elimination device: subtracting successive equations removes the common a term and genuinely reduces the number of equations and unknowns.
The speaker explicitly states that nothing is being squared in the notation .
Thinking that means squaring the value of .
The superscript 2 indicates applying the difference operator twice, not squaring the result.
The speaker warns that the formula only applies when the data points have x-values 0,1,...,n.
One might think the displayed formula works for arbitrary x-values.
The video explicitly restricts this version of the formula to equally spaced nodes ,1,...,n and says a later generalization handles other cases.
The speaker says we never actually multiply by x-n; we stop at x-(n-1).
The last displayed product ends with (x-(n-1)).
Because the data run up to , one may expect the last product to include x-n.
The final falling-factor product stops at x-(n-1); the role of the nth data point is already contained in computing .
After posing the smooth-curve question, the speaker says they will use a polynomial instead.
The board edits "smooth curve" into "polynomial".
The general interpolation problem is specialized by choosing polynomials as the class of admissible approximating curves.
The speaker moves from saying they will use a polynomial to stating the degree needed for n points.
The board extends "polynomial" to "polynomial of degree n-1".
Once polynomials are chosen, the video adds the specific rule that n points correspond to degree n-1.
The speaker says they will lead to the general formula through an example and then introduces four data points and a cubic polynomial.
The example board uses exactly four points and a cubic, matching the earlier case.
The cubic-through-four-points example is the instance of the general degree rule stated just before the scene change.
The displayed cubic and the displayed four equations appear together on the same example board.
The algebraic system is derived directly from the cubic setup by enforcing passage through the four specified points.
The goal statement with the cubic and four points is immediately followed by the displayed four-equation system.
The narrator explains that substituting the points into the cubic yields the system.
The interpolation definition is applied pointwise to generate the concrete linear system shown on the board.
The narrator says the new system has a b in each equation, so the same thing can probably be done again to kill off yet another variable.
The reduced equations visibly all begin with a b term.
The second-difference equations themselves are not written in this clip.
The success of eliminating a by first differences suggests the same recursive strategy can be applied to the reduced system to eliminate b next.
The speaker explains how finding allows solving for d, and then working backwards to find c, b, and a.
The method of finite differences is applied to the system of equations derived from the cubic polynomial to solve for its coefficients.
The speaker says the patterns in the four-point equation can perhaps be generalized to n points.
The second board presents the -point formula as the direct extension.
The generalized formula extends the four-point Newton forward interpolation expression to equally spaced points.
The coefficient formulas are solved first, then substituted to produce the boxed Newton form.
The four-point Newton formula depends on first expressing a,b,c,d in terms of forward differences.
The derivation begins from +bx+cx^ and then rewrites its coefficients.
The coefficient expressions are meaningful only relative to the initial cubic template.
The speaker describes the predictable term pattern immediately before writing the generalized formula.
The structural pattern explanation is the rule used to read and construct the generalized formula.
Title and problem statement are visible on the overview board.
Speaker says they will not use just any old smooth curve and will use a polynomial.
Speaker states the general rule for n points.
Board writes "polynomial of degree n-1".
Example board shows the cubic and the four-point system.
Speaker says the problem is impossible to answer with certainty.
The board title and goal statement explicitly ask for a cubic through four points.
The narrator reads the goal aloud.
The four displayed equations correspond to the four interpolation points.
The narrator explains that plugging in the points gives a system of four equations with four unknowns.
The narrator substitutes and says the last three terms disappear.
The first equation on the board is .
The narrator says Newton observed each row has a single a term and took differences to kill the a terms.
The board shows the three Delta equations derived from successive subtraction.
The narrator explicitly calls the Delta y terms first finite differences.
The board labels them Delta , Delta , Delta .
The narrator says the same thing can probably be done to the new system to kill off yet another variable.
Each reduced equation contains a b term.
The next elimination step is announced but not executed in this segment.
Speaker explains taking finite differences to kill off the 'b' term.
Covered · Overview section: problem statement, motivation, warning about uncertainty, replacement of smooth curve by polynomial, and degree pattern illustrated on the graph.
Covered · Brief transition between the overview board and the example board.
Covered · Worked example setup: cubic form, four specified points, and the resulting linear system; the clip ends before solving it.
Covered · Problem statement and cubic form are introduced.
Covered · Substitution of the four points produces the 4-equation linear system.
Covered · Narrator contrasts modern linear algebra with Newton’s historical method.
Covered · Successive subtraction defines Delta , Delta , Delta and eliminates a.
Covered · The reduced 3-by-3 system is identified, and the narrator proposes repeating the method; the actual second-difference computation is not shown in this clip.
Covered · The entire segment covers the explanation and derivation of using finite differences to solve for the coefficients of a cubic polynomial.
Covered · Opening board shows the cubic template and the solved coefficient formulas in terms of forward differences.
Covered · The presenter writes and explains the boxed four-point Newton interpolation formula.
Covered · The board changes to the generalized -point formula and the speaker describes the repeating term pattern.
Covered · The presenter gives two remarks: the formula requires x-values 0 through n, and the final product stops at x-(n-1).
Reviewed English material v1: card "Polynomial interpolation problem" at 0:00 — The video poses the basic question: given a finite set of data points, can we find a curve passing through them so as to approximate an unknown function between the samples? This is presented as the motivating problem before any formula is introduced.
Continue with related videos in the same subject.
YouTube8:13
YouTube6:03
YouTube5:16