Numerical Analysis 7.4. Gaussian Quadrature: Efficient Integration

Описание: The video provided is a lecture on Gaussian Quadrature, specifically focusing on Section 7.4 of a Numerical Analysis course. It explains how to approximate integrals more efficiently than standard Newton-Cotes formulas.

Core Concepts of Gaussian Quadrature
The Objective: Unlike previous methods where points were fixed, Gaussian quadrature seeks to find both the weights (c
i
​
) and the mesh points (x
i
​
) that make the approximation exact for polynomials of the highest possible degree [00:44].

Degree of Exactness: For an n-point formula, there are 2n parameters to determine. This allows the formula to be exact for polynomials up to degree 2n−1 [03:37].

Nonlinear System: Finding these parameters involves solving a system of nonlinear algebraic equations derived from the exact integration of monomial functions (x
0
,x
1
,...,x
2n−1
) [05:15].

Legendre Polynomials and Orthogonality
Orthogonal Functions: The lecture introduces the concept of function orthogonality, where the integral of the product of two functions over an interval is zero [14:40].

Legendre Polynomials (P
i
​
): These are a sequence of polynomials where each P
i
​
(of degree i) is orthogonal to all polynomials of a lower degree [21:37].

Determining Mesh Points: Crucially, the mesh points (x
i
​
) for a Gaussian quadrature formula are the roots of the Legendre polynomial of degree n [24:28]. These roots are always distinct, real, and located within the interval (−1,1) [23:22].

Practical Application and Error
The 2-Point Formula (n=2): On the interval [−1,1], the weights are both 1, and the points are ±
3
3

​

​
[12:02]. An example calculating the integral of e
x
demonstrates that even with just two points, the error is very small (0.007) [13:41].

Efficiency: The error in Gaussian quadrature converges to zero at an exponential speed as n increases, which is much faster than the polynomial speed of Newton-Cotes formulas [28:44].

Arbitrary Intervals: To use this method on any interval [a,b], a linear change of variables is used to transform the integral into one over the standard interval [−1,1] [30:14].

The lecture concludes with a numerical example calculating ∫
0
1
​
x
2
e
x
dx using a 2-point Gaussian formula, showing a high level of accuracy with minimal computation [31:37].