Jim Lambers MAT 460/560 Fall Semester 2009-10 Lecture 18 Notes These notes correspond to Section 3.2 in the text. Interpolation Using Equally Spaced Points Suppose that the interpolation points π₯0 , π₯1 , . . . , π₯π are equally spaced; that is, π₯π = π₯0 + πβ for some positive number β. In this case, the Newton interpolating polynomial can be simpliο¬ed, since the denominators of all of the divided diο¬erences can be expressed in terms of the spacing β. If we recall the forward diο¬erence operator Δ, deο¬ned by Δπ₯π = π₯π+1 − π₯π , where {π₯π } is any sequence, then the divided diο¬erences π [π₯0 , π₯1 , . . . , π₯π ] are given by π [π₯0 , π₯1 , . . . , π₯π ] = 1 Δπ π (π₯0 ). π!βπ The interpolating polynomial can then be described by the Newton forward-diο¬erence formula ) π ( ∑ π Δπ π (π₯0 ), ππ (π₯) = π [π₯0 ] + π π=1 where the new variable π is related to π₯ by π = ( and the extended binomial coeο¬cient ( π π ) = π π π₯ − π₯0 , β ) is deο¬ned by π (π − 1)(π − 2) ⋅ ⋅ ⋅ (π − π + 1) , π! where π is a nonnegative integer. Example We will use the Newton forward-diο¬erence formula ) π ( ∑ π ππ (π₯) = π [π₯0 ] + Δπ π (π₯0 ) π π=1 to compute the interpolating polynomial π3 (π₯) that ο¬ts the data 1 π 0 1 2 3 π₯π −1 0 1 2 π (π₯π ) 3 −4 5 −6 In other words, we must have π3 (−1) = 3, π3 (0) = −4, π3 (1) = 5, and π3 (2) = −6. Note that the interpolation points π₯0 = −1, π₯1 = 0, π₯2 = 1 and π₯3 = 2 are equally spaced, with spacing β = 1. To apply the forward-diο¬erence formula, we deο¬ne π = (π₯ − π₯0 )/β = π₯ + 1 and compute ( ) π = π 1 = π₯ + 1, ( ) π (π − 1) π = 2 2 π₯(π₯ + 1) = , 2 ( ) π (π − 1)(π − 2) π = 3 6 (π₯ + 1)π₯(π₯ − 1) , = 6 π [π₯0 ] = π (π₯0 ) = 3, Δπ (π₯0 ) = π (π₯1 ) − π (π₯0 ) = −4 − 3 = −7, 2 Δ π (π₯0 ) = Δ(Δπ (π₯0 )) = Δ[π (π₯1 ) − π (π₯0 )] = [π (π₯2 ) − π (π₯1 )] − [π (π₯1 ) − π (π₯0 )] = π (π₯2 ) − 2π (π₯1 ) + π (π₯0 ) = 5 − 2(−4) + 3, = 16 3 Δ π (π₯0 ) = Δ(Δ2 π (π₯0 )) = Δ[π (π₯2 ) − 2π (π₯1 ) + π (π₯0 )] = [π (π₯3 ) − π (π₯2 )] − 2[π (π₯2 ) − π (π₯1 )] + [π (π₯1 ) − π (π₯0 )] = π (π₯3 ) − 3π (π₯2 ) + 3π (π₯1 ) − π (π₯0 ) = −6 − 3(5) + 3(−4) − 3 2 = −36. It follows that π3 (π₯) = = = = ) 3 ( ∑ π π [π₯0 ] + Δπ π (π₯0 ) π π=1 ( ) ( ) ( ) π π π 2 3+ Δπ (π₯0 ) + Δ π (π₯0 ) + Δ3 π (π₯0 ) 1 1 2 (π₯ + 1)π₯(π₯ − 1) π₯(π₯ + 1) 16 + (−36) 3 + (π₯ + 1)(−7) + 2 6 3 − 7(π₯ + 1) + 8(π₯ + 1)π₯ − 6(π₯ + 1)π₯(π₯ − 1). Note that the forward-diο¬erence formula computes the same form of the interpolating polynomial as the Newton divided-diο¬erence formula. β‘ If we deο¬ne the backward diο¬erence operator ∇ by ∇π₯π = π₯π − π₯π−1 , for any sequence {π₯π }, then we obtain the Newton backward-diο¬erence formula ππ (π₯) = π [π₯π ] + π ∑ π ( (−1) π=1 −π π ) ∇π π (π₯π ), where π = (π₯ − π₯π )/β, and the preceding deο¬nition of the extended binomial coeο¬cient applies. Example We will use the Newton backward-diο¬erence formula ππ (π₯) = π [π₯π ] + π ∑ π (−1) π=1 ( −π π ) ∇π π (π₯π ) to compute the interpolating polynomial π3 (π₯) that ο¬ts the data π 0 1 2 3 π₯π −1 0 1 2 π (π₯π ) 3 −4 5 −6 In other words, we must have π3 (−1) = 3, π3 (0) = −4, π3 (1) = 5, and π3 (2) = −6. Note that the interpolation points π₯0 = −1, π₯1 = 0, π₯2 = 1 and π₯3 = 2 are equally spaced, with spacing β = 1. 3 To apply the backward-diο¬erence formula, we deο¬ne π = (π₯ − π₯3 )/β = π₯ − 2 and compute ( ) −π = −π 1 = −(π₯ − 2), ( ) −π (−π − 1) −π = 2 2 π (π + 1) = 2 (π₯ − 2)(π₯ − 1) = , 2 ( ) −π (−π − 1)(−π − 2) −π = 3 6 π (π + 1)(π + 2) = − 6 (π₯ − 2)(π₯ − 1)π₯ , = − 6 π [π₯3 ] = π (π₯3 ) = −6, ∇π (π₯3 ) = π (π₯3 ) − π (π₯2 ) = −6 − 5 = −11, 2 ∇ π (π₯3 ) = ∇(∇π (π₯3 )) = Δ[π (π₯3 ) − π (π₯2 )] = [π (π₯3 ) − π (π₯2 )] − [π (π₯ − 2) − π (π₯1 )] = π (π₯3 ) − 2π (π₯2 ) + π (π₯1 ) = −6 − 2(5) + −4, = −20 3 ∇ π (π₯3 ) = ∇(∇2 π (π₯3 )) = ∇[π (π₯3 ) − 2π (π₯2 ) + π (π₯1 )] = [π (π₯3 ) − π (π₯2 )] − 2[π (π₯2 ) − π (π₯1 )] + [π (π₯1 ) − π (π₯0 )] = π (π₯3 ) − 3π (π₯2 ) + 3π (π₯1 ) − π (π₯0 ) = −6 − 3(5) + 3(−4) − 3 = −36. 4 It follows that π3 (π₯) = π [π₯3 ] + 3 ∑ (−1)π π=1 ( ( −π π ∇π π (π₯3 ) ) ( ) −π −π 2 = −6 − ∇π (π₯3 ) + ∇ π (π₯3 ) − ∇3 π (π₯3 ) 1 2 −(π₯ − 2)(π₯ − 1)π₯ (π₯ − 2)(π₯ − 1) (−20) − (−36) = −6 − [−(π₯ − 2)](−11) + 2 6 = −6 − 11(π₯ − 2) − 10(π₯ − 2)(π₯ − 1) − 6(π₯ − 2)(π₯ − 1)π₯. −π 1 ) ) ( Note that the backward-diο¬erence formula does not compute the same form of the interpolating polynomial ππ (π₯) as the Newton divided-diο¬erence formula. Instead, it computes a diο¬erent Newton form of ππ (π₯) given by ππ (π₯) = π ∑ π=0 π [π₯π , π₯π+1 , . . . , π₯π ] π ∏ (π₯ − π₯π ) π=π+1 = π [π₯π ] + π [π₯π−1 , π₯π ](π₯ − π₯π ) + π [π₯π−2 , π₯π−1 , π₯π ](π₯ − π₯π−1 )(π₯ − π₯π ) + ⋅ ⋅ ⋅ + π [π₯0 , π₯1 , . . . , π₯π ](π₯ − π₯1 )(π₯ − π₯2 ) ⋅ ⋅ ⋅ (π₯ − π₯π−1 )(π₯ − π₯π ). The divided-diο¬erences π [π₯π , . . . , π₯π ], for π = 0, 1, . . . , π, can be obtained by constructing a divideddiο¬erence table as in the ο¬rst example. These divided diο¬erences appear in the bottom row of the table, whereas the divided diο¬erences used in the forward-diο¬erence formula appear in the top row. β‘ 5