📚 The Derivative in Any Direction | 任意方向的方向导数
The derivative of a function of several variables measures the rate of change along a specific direction. While partial derivatives only describe change along coordinate axes, the directional derivative extends this concept to any direction in the domain. It is a fundamental tool in multivariable calculus, used in optimization, physics, and engineering.
多元函数的导数衡量函数沿特定方向的变化率。虽然偏导数只能描述沿坐标轴方向的变化，但方向导数将这一概念扩展到定义域内的任意方向。它是多变量微积分中的基本工具，应用于最优化、物理和工程等领域。
1. Definition of the Directional Derivative | 方向导数的定义
u₁, b+hu₂) − f(a,b)] / h, provided the limit exists. The vector u must have length 1 to correctly measure the rate of change.
对于函数 f(x,y) 在点 (a,b) 处，沿单位向量 u = (u₁, u₂) 的方向导数定义为 Dᵤ f(a,b) = limₕ→₀ [f(a+hu₁, b+hu₂) − f(a,b)] / h，假设该极限存在。向量 u 必须长度为 1，才能正确衡量变化率。
2. Geometric Interpretation | 几何意义
a vertical plane parallel to the direction vector u. It tells us how steeply the function rises or falls if we move from (a,b) in that specific direction, exactly like the slope of a single-variable function along a cross section.
几何上，方向导数表示曲面 z = f(x,y) 与平行于方向向量 u 的竖直平面相交所形成曲线的切线斜率。它告诉我们如果从点 (a,b) 沿该特定方向移动，函数值升高的或下降的陡峭程度，就像单变量函数沿截面的斜率一样。
3. Relation to Partial Derivatives | 与偏导数的关系
rectional derivatives generalize partial derivatives, which are simply the derivatives in the coordinate axis directions. All partial derivatives are special cases of directional derivatives.
如果 u 指向 x 轴正方向，u = (1,0)，则 Dᵤ f = ∂f/∂x，即对 x 的偏导数。类似地，u = (0,1) 给出 ∂f/∂y。因此，方向导数推广了偏导数，偏导数就是沿坐标轴方向的方向导数。所有偏导数都是方向导数的特例。
4. Gradient and the Directional Derivative | 梯度与方向导数
When f is differentiable, the directional derivative can be expressed as the dot product of the gradient and the direction vector:
Dᵤ f = ∇f · u
The gradient ∇ nts in the direction of the greatest rate of increase. This compact formula makes calculation straightforward and reveals the intrinsic link between gradient and directional change.
当 f 可微时，方向导数可表示为梯度与方向向量的点积：
Dᵤ f = ∇f · u
梯度 ∇f = (∂f/∂x, ∂f/∂y) 是一个向量，指向函数值增加最快的方向。这个简洁的公式使计算变得简单，并揭示了梯度与方向变化之间的内在联系。
5. Computing the Directional Derivative | 方向导数的计算
To find Dᵤ f at (a,b), first compute the gradient ∇ direction vector is a unit vector; if a non-unit vector v is given, normalize it: u = v / |v|. The result is a scalar representing the rate of change in that direction.
计算 (a,b) 处的 Dᵤ f：先求梯度 ∇f(a,b)，再点乘单位向量 u。务必确保方向向量是单位向量；若给定非单位向量 v，需将其单位化：u = v / |v|。结果是一个标量，表示该方向的变化率。
6. Direction of Maximum Increase | 最大上升方向
nal derivative is maximized when u points in the same direction as the gradient, i.e., u = ∇f / |∇f|. The maximum value is |∇f| t negative) directional derivative occurs when u is opposite to the gradient, with value −|∇f|. This property is crucial in gradient ascent methods.
当 u 指向梯度方向，即 u = ∇f / |∇f| 时，方向导数取得最大值 |∇f|。反之，当 u 与梯度方向相反时，取得最小值（最负）−|∇f|。这性质在梯度上升方法中至关重要。
7. Direction of Zero Change | 变化率为零的方向
The directional derivative is zero in directions orthogonal to the gradient. At a point, the level curve of f has a tangent perpendicular to the gradient; moving along the level curve gives zero instantaneous change. This fact is used to find equations of tangent lines to contours and to understand function behaviour near critical points.
方向导数在垂直于梯度的方向上为零。在某一点处，f 的等值线的切线与梯度垂直；沿着等值线移动时瞬时变化率为零。这一事实用于求等高线的切线方程，并理解临界点附近的函数行为。
8. Applications – Tangent Planes and Linear Approximation | 应用 – 切平面与线性近似
ear approximation for a small step h in direction u: f(a+hu₁, b+hu₂) ≈ f(a,b) + h Dᵤ f(a,b).
方向导数是切平面方程的基础：z = f(a,b) + fₓ(a,b)(x−a) + fᵧ(a,b)(y−b)。这个线性化利用偏导数（沿坐标轴的方向导数）来近似 (a,b) 附近的函数值。这一概念可推广至方向增量，给出沿方向 u 移动小步长 h 的线性近似：f(a+hu₁, b+hu₂) ≈ f(a,b) + h Dᵤ f(a,b)。
9. Example Walkthrough | 实例演练
+ y² at point (1,2). Find the directional derivative in the direction of v = (3,4). First, ∇f = (2x, 2y), so ∇f(1,2) = (2,4). Next, unit vector u = (3,4)/√3²+4² ). Then Dᵤ f = (2,4)·(3/5, 4/5) = 6/5 + 16/5 = 22/5 = 4.4. The maximum rate of increase is |∇f| = √2²+4² = √20 ≈ 4.47, occurring in direction (2,4).
考虑 f(x,y) = x² + y² 在点 (1,2)。求沿 v = (3,4) 方向的方向导数。首先，∇f = (2x, 2y)，所以 ∇f(1,2) = (2,4)。然后，单位向量 u = (3,4)/√3²+4² = (3/5, 4/5)。故 Dᵤ f = (2,4)·(3/5, 4/5) = 6/5 + 16/5 = 22/5 = 4.4。最大增长率为 |∇f| = √2²+4² = √20 ≈ 4.47，发生在 (2,4) 方向。
10. Extension to Three Variables | 推广到三元函数
e directional derivative at (a,b,c) in direction of unit vector u = (u₁, u₂, u₃) is Dᵤ f = ∇f · u, where ∇f = (fₓ, fᵧ, fz). The geometric interpretation involves a hyper-surface in four dimensions, but the computational rule remains the same. The gradient points toward the steepest increase in the input space, and its magnitude gives the maximum rate.
对于 f(x,y,z)，在 (a,b,c) 处沿单位向量 u = (u₁, u₂, u₃) 的方向导数为 Dᵤ f = ∇f · u，其中 ∇f = (fₓ, fᵧ, fz)。几何意义涉及四维空间中的超曲面，但计算规则保持不变。梯度在自变量空间中指向最陡上升方向，其大小给出最大变化率。
11. Importance in Multivariable Optimization | 在多变量最优化中的重要性
In constrained or unconstrained optimization, gradients guide algorithms like gradient descent. The directional derivative quantifies how the objective function changes along a search direction, helping to find minima or maxima efficiently. Understanding directional derivatives is essential for mastering Lagrange multipliers, sensitivity analysis, and machine learning cost functions.
在约束或无约束最优化中，梯度引导梯度下降等算法。方向导数量化了目标函数沿搜索方向的变化情况，有助于高效寻找极小值或极大值。理解方向导数对于掌握拉格朗日乘数法、敏感性分析和机器学习成本函数至关重要。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图