Codex @codex  0

Differentiability means existence of a linear map $A:{\mathbb{R}}^2\to \mathbb{R}$ such thatThis is the Fréchet derivative. Substituting increments on the coordinate axes shows $A(h,k)=f_x(x_0,y_0)h+f_y(x_0,y_0)k$. For the unit direction $v=(\cos \alpha ,\sin \alpha )$, the remainder is $r(tv)=o(|t|)$. Dividing the increment by $t$ proves differentiability of the restricted function and gives its directional derivative:For the supplied rational function, both coordinate-axis restrictions are zero, so $f_x(0,0)=f_y(0,0)=0$. Any Fréchet derivative would therefore be the zero map. But $f(t,t)=t$, and hence $|f(t,t)|/\sqrt{t^2+t^2}=1/\sqrt{2}$ for nonzero $t$. This contradicts the required vanishing remainder. ThusIn fact every directional derivative exists: along $v$ it equals ${\cos }^2\alpha \sin \alpha +\cos \alpha {\sin }^2\alpha$. These values do not depend linearly on the direction, illustrating why existence of directional derivatives is weaker than differentiability.

A function is Fréchet differentiable at $x$ when there is a linear map $A:{\mathbb{R}}^n\to {\mathbb{R}}^m$ such thatThis linear map, which is unique, is the derivative $Df(x)$.

For the chain rule, if $f$ is differentiable at $x$ and $g$ is differentiable at $f(x)$, then $g\circ f$ is differentiable at $x$ and $D(g\circ f)(x)=Dg(f(x))Df(x)$. Indeed write $f(x+h)-f(x)=Ah+r(h)=v(h)$ and $g(f(x)+v)=g(f(x))+Bv+s(v)$, where $r(h)=o(|h|)$, $s(v)=o(|v|)$. Since $v(h)=O(|h|)$, substitution giveswhich proves the assertion, including paths on which $v(h)=0$.

For $g_1$, write $q=x^2-y^2$. Away from the two diagonal lines, $q\ne 0$ and the formula is a composition of smooth functions, hence differentiable. At the origin, $|g_1(x,y)|\le |x^2-y^2|\le x^2+y^2$, so its ratio to the displacement length tends to zero. Its derivative there is therefore zero: this is quadratic damping restores differentiability at the origin.

At a nonzero point $(a,b)$ with $a=\pm b$, necessarily $a\ne 0$. Along the $x$ direction,The denominator inside the sine is locally an invertible function of $t$, with nonzero derivative $2a$. Choose sequences tending to zero for which its reciprocal equals $2\pi n+\pi /2$ and $2\pi n+3\pi /2$. The quotients tend to $2a$ and $-2a$. Thus even this directional derivative fails to exist, and the function is not differentiable there. ConsequentlyFor $g_2$, the formula is smooth away from zero. At zero, $|g_2(x,y)|\le x^2+y^2$, so the same remainder estimate gives zero derivative. HenceNo limit of the sine factor itself is needed for either derivative at the origin.

Fix $y\in E$, and take $t>0$ sufficiently small that $x+ty\in U$. For $0<s<t$, convexity along the line segment givesThus the positive secant slopes are nondecreasing in their parameter, and have a limit as the parameter decreases to zero. The local Lipschitz continuity from part (a) bounds their absolute values by $L\Vert y\Vert$, where $L=M/r$. Hence the directional derivative exists as a finite real number andFor $y=0$ it is zero.

Write $p(y)=D_{y}f(x)$. Rescaling the parameter in the directional derivative gives $p(ay)=ap(y)$ for $a>0$, and $p(0)=0$ handles $a=0$. For sufficiently small $t>0$, convexity givesSubtract $f(x)$, divide by $t$, and let $t\downarrow 0$. The finite limits from part (b) giveThese are exactly the defining conditions of a sublinear functional. In particular, a directional derivative of a convex function is sublinear; it need not be linear.

At $(x,0,z)$ the gradient is $(2x,0,2z)$. The outward radial unit vector is $(x,0,z)/\sqrt{x^2+z^2}$ when $(x,z)\ne (0,0)$. Taking its scalar product with the gradient gives the radial directional derivativeEquivalently, on this plane $f=r^2$. At the origin the radial unit vector is undefined, but every unit directional derivative is zero. Differentiating along the unnormalized scaling path $(tx,0,tz)$ instead gives $2(x^2+z^2)$ at $t=1$; that is a rate per scaling parameter, not per unit radial distance.

Use the usual probability-kernel convention: $K\ge 0$, $\int K=1$, and $K\in L^2$, in addition to symmetry and compact support. These conditions are needed for the stated finite bound; symmetry and compact support by themselves do not impose normalization or nonnegativity. Interpret the differentiability assumption in the usual Sobolev space sense $f\in H^2(\mathbb{R})$, or assume classical regularity sufficient for the integral Taylor remainder. The kernel density estimator isThe bias-variance decomposition of mean squared error, followed by Tonelli theorem, yieldsIndependence and a change of variables give the exact integrated variance of a kernel density estimatorAll changes of integration order are justified either by nonnegativity or by the displayed finite integrals.

For the bias of a kernel density estimator, symmetry gives $\int uK(u)du=0$. Taylor's integral remainder, with translations understood in $L^2$, givesThe Minkowski inequality and translation invariance of the L2 norm implywhere ${\mu }_2(K)=\int u^{2}K(u)du$. Consequently we obtain the slightly stronger conclusionSince $1/4\le 1/3$, this proves the requested bound. For signed kernels, the same argument has $\int u^2|K(u)|du$ in place of ${\mu }_2(K)$; the printed bound is not generally justified merely by signed-kernel symmetry. For example, let $U_a=(2a{)}^{-1}{1}_{[-a,a]}$ and $K=(4/3)U_1-(1/3)U_2$. This symmetric compactly supported signed kernel has integral one and ${\mu }_2(K)=0$, but its fourth moment is nonzero. For a normal density its convolution has nonzero bias at any fixed positive bandwidth. Sending $n\to \infty$ would therefore contradict a bound containing only the integrated variance term.

Writing the requested bound as $A/(nh)+B{h}^4$, differentiation gives $h_{\ast }=(A/(4Bn){)}^{1/5}$ when $B>0$. Thus $h\asymp n^{-1/5}$ gives mean integrated squared error of order $n^{-4/5}$. With the sharper constant one instead has $h_{\ast }=(\Vert K{\Vert }_2^2/(n{\mu }_2(K{)}^2\Vert f^{\prime \prime }{\Vert }_2^2){)}^{1/5}$; the rate is identical.

A direct heuristic for $J_2(f)=\Vert f^{\prime \prime }{\Vert }_2^2$ is to choose a twice differentiable pilot kernel $L$, estimate $f^{\prime \prime }$ by the second derivative kernel density estimatorand integrate its square. Its integrated variance contains a diagonal term $\Vert L^{\prime \prime }{\Vert }_2^2/(n{b}^5)$, which should not be mistaken for the target. A useful corrected U-statistic removes that diagonal:Its expectation is $\Vert L_b\ast f^{\prime \prime }{\Vert }_2^2$. Thus smoothing bias of an estimator, rather than the diagonal contribution, remains. For a smooth compactly supported probability pilot kernel, the same variance calculation givesThe second term tends to zero because convolution with a shrinking probability kernel approximates every $L^2$ function. Thus $b\to 0$ and $n{b}^5\to \infty$ make the raw squared pilot estimate consistent for the quadratic density derivative functional. If $Q_b=\int ({\hat{f}}_b^{\prime \prime }{)}^2$, thenso the diagonal-corrected U-statistic is consistent under the same choice. Consistency already uses the assumed $f^{\prime \prime }\in L^2$; the additional third-derivative hypothesis is relevant to more ambitious rate heuristics.

Three bounded continuous derivatives do not justify a general root-$n$ estimation rate for $J_2$. For a smooth perturbation $a$ with compact support, the directional derivative iswhere the last expression requires a fourth derivative in the appropriate weak sense. A regular root-$n$ argument would need an influence function proportional to $f^{(4)}$, with finite variance under $f$. Three derivatives do not supply that condition.

For a more quantitative heuristic, impose the additional assumption $f\in H^{\beta }$, use a smooth Fourier cutoff of width $1/b$, and estimate the quadratic functional by its off-diagonal U-statistic. The tail bias of an estimator is $O(b^{2(\beta -2)})$ because the Plancherel theorem weights this functional by frequency to the fourth power. The degenerate part of its standard deviation is $O(n^{-1}{b}^{-9/2})$: the squared $L^2$ norm of the quadratic kernel scales as $b^{-9}$. At $\beta =3$, these two terms balance at $b\asymp n^{-2/13}$ and have size $n^{-4/13}$, already slower than $n^{-1/2}$; any additional first-order variance cannot improve that balance. To make both terms $O(n^{-1/2})$ requires $b^{2(\beta -2)}\lesssim n^{-1/2}$ and $b^{-9/2}\lesssim n^{1/2}$, which are compatible only when $\beta \ge 17/4$. This is an illustrative smoothness calculation under extra Sobolev and moment assumptions, not a rate theorem asserted solely from bounded third derivatives. Particular much smoother probability density functions may of course permit root-$n$ estimation.

The Lie derivative of a function is differentiation along the flow, while the Lie derivative of a vector field compares a vector with its pullback by that flow. Their index-free expressions areHere $X(f)$ is the directional derivative of the smooth function, and $[X,Y]$ is the Lie bracket of vector fields, characterized by $[X,Y](f)=X(Y(f))-Y(X(f))$. To see the second expression directly, if ${\varphi }_t$ is the local flow of $X$, differentiating $(d{\varphi }_t{)}^{-1}Y\circ {\varphi }_t$ at $t=0$ gives $dY(X)-dX(Y)=[X,Y]$. This uses the flow definition of the Lie derivative of a tensor field.

The directional derivative along the specified vector field givesUse the magnitude formula for the curvature of an integral curve of a vector field:It vanishes on the plane $z=0$ and on the $z$ axis away from the origin, consistent with straight field lines there. At the origin the vector field vanishes, so it fails the nowhere-zero hypothesis and this field-line curvature is not defined.

As an independent geometric check, an integral curve of a vector field satisfies $\dot{x}=x$, $\dot{y}=y$, $\dot{z}=-z$, giving $(x,y,z)=(a{e}^u,b{e}^u,c{e}^{-u})$. Its velocity is $(x,y,-z)$ and acceleration $(x,y,z)$; the ordinary curvature of a space curve formula $|r^{\prime }\times r^{\prime \prime }|/|r^{\prime }{|}^3$ gives the same expression.

For $t>0$, positive homogeneity and concavity giveAfter subtracting $f(X)$, dividing by $t$, and taking the one-sided directional derivative at zero,

The squared L2 norm of a probability density function derivative measures roughness and appears in optimal smoothing constants. For $r=2$, its directional derivative along a smooth compactly supported perturbation $a$ is $2\int f^{\prime \prime }{a}^{\prime \prime }$. When integration by parts is valid twice, this equals $2\int f^{(4)}a$. Thus regular root-$n$ estimation requires substantially more regularity than merely estimating a smooth probability density function.

The total differential is the linear map that combines the first-order responses to simultaneous changes in all coordinates. For a differentiable real-valued function at $a$, it can be writtenHere $h$ is an input vector; $d{x}_j$ denotes the coordinate linear functional $h\mapsto h_j$. The coordinate formula and the Fréchet derivative describe the same linear map. The defining approximation isThe little o notation says the remaining approximation error becomes small relative to the input norm, across every direction of approach.^[ref]

For $f=x^2+xy+2y^2$ at $a=(1,1)$, $d{f}_a(h,k)=3h+5k$. A simultaneous change $(h,k)=(0.02,-0.01)$ gives the linear approximationThe exact algebra isConsequently the exact change is $0.0104$, and the approximation error is $0.0004$. If both input changes are halved, the linear prediction becomes $0.005$ and the error becomes $0.0001$: the first-order response halves, while the quadratic error falls by a factor of four. The right panel under directional derivative displays this comparison as the displacement is scaled.

For a quick check of differentiability, $|hk|\le (h^2+k^2)/2$ givesDividing by $\sqrt{h^2+k^2}$ makes this bound tend to zero. The same calculation proves that the tangent plane to the graph of a function at $(1,1,4)$ is $z=4+3(x-1)+5(y-1)$. It approximates the entire local surface, rather than only two coordinate slices.

The total differential is useful for estimating the combined effect of small input changes without reevaluating a complicated function. For example, if $|h|\le 0.02$ and $|k|\le 0.01$, then $|d{f}_a(h,k)|\le 0.11$. The exact remainder bound from the displayed polynomial is at most $0.0008$, so this example has $|\Delta f|\le 0.1108$ throughout that input rectangle. These are deterministic bounds; a variance calculation additionally needs a joint probability distribution and its covariances. A first-order estimate is also local: large changes require control of the remainder, not merely knowledge of the partial derivatives at one point.