More equally spaced samples need not improve high-degree polynomial interpolation: Runge’s function develops large endpoint oscillations. Chebyshev nodes place more samples near both ends. For this analytic function, that redistribution controls the endpoint problem and produces convergent polynomial approximations.高次多项式插值中,等距样本增加不一定让近似更准:龙格函数的插值曲线会在区间两端产生明显振荡。切比雪夫节点将更多样本布置在两端。对于图中这样的解析函数,这种重新分配能抑制端部问题,使插值逐步收敛。
Green is the exact function. Gold uses equal nodes; cyan uses Chebyshev nodes, shown in matching rows below. Degrees 6, 8 and 10 share one unclipped scale. Moving markers compare both approximations with the true value, especially near the endpoints.绿线是真实函数,金线使用等距节点,青线使用切比雪夫节点;底部两排同色圆点列出节点位置。6、8、10 次插值共用同一尺度,曲线没有裁剪。移动标记比较两种插值与真值的偏差,尤其留意区间两端。