--- title: Cauchy Density in Categories & Probability url: https://www.emergentmind.com/topics/cauchy-density type: topic --- # Cauchy Density in Categories & Probability Searching arXiv for the cited papers and closely related context. arXiv search query: id:2507.07869 OR title:"Cauchy density" “Cauchy density” appears in the cited literature in two distinct senses. In \(\mathscr V\)-enriched category theory, it names a condition on a \(\mathscr V\)-functor, introduced by Lawvere in the same paper that defined the Cauchy completion of a \(\mathscr V\)-category; in that setting, the condition makes the functor analogous to a map of metric spaces whose image is topologically dense in its codomain [2507.07869]. In probability theory, it denotes the probability density function of the Cauchy distribution, \[ f(x;x_0,\gamma)=\frac{1}{\pi\,\gamma\,[1+((x-x_0)/\gamma)^2]}, \] with location parameter \(x_0\in\mathbb R\) and scale \(\gamma>0\) [2407.09583]. The two usages are mathematically unrelated at the level of formal definition, but both are tied to classical notions of completion, boundary behavior, and density. ## 1. Terminological scope and basic forms In the enriched-categorical sense, let \(\mathscr V\) be a complete and cocomplete symmetric monoidal closed category. For small \(\mathscr V\)-categories \(\mathscr A,\mathscr B\), a profunctor \(\phi:\mathscr A\pto\mathscr B\) is a \(\mathscr V\)-functor \(\phi:\mathscr B^{op}\otimes\mathscr A\to\mathscr V\), and every \(\mathscr V\)-functor \(F:\mathscr A\to\mathscr B\) gives a pair of adjoint profunctors \[ F_*:\mathscr A\pto\mathscr B,\qquad F^*:\mathscr B\pto\mathscr A, \] with \[ F_*(b,a)=\mathscr B(b,Fa),\qquad F^*(a,b)=\mathscr B(Fa,b), \] and \(F_*\dashv F^*\). The functor \(F\) is called Cauchy dense if the counit \[ \eps^F:F_*\otimes F^*\longrightarrow \mathscr B(-,-) \] is an isomorphism of profunctors [2507.07869]. In the probabilistic sense, the Cauchy probability density function is \[ f(x;x_0,\gamma)=\frac{1}{\pi\,\gamma\,[1+((x-x_0)/\gamma)^2]}. \] A special case is the standard Cauchy \(C(0,1)\), obtained when \(x_0=0\) and \(\gamma=1\) [2407.09583]. | Sense | Object | Defining form | |---|---|---| | Enriched category theory | \(\mathscr V\)-functor \(F:\mathscr A\to\mathscr B\) | \(\eps^F:F_*\otimes F^*\to\mathscr B(-,-)\) is an isomorphism | | Probability theory | Density on \(\mathbb R\) | \(f(x;x_0,\gamma)=\frac{1}{\pi\gamma[1+((x-x_0)/\gamma)^2]}\) | A plausible implication is that the shared terminology reflects a common historical association with completion and limiting behavior, even though the formal frameworks are different. ## 2. Cauchy density in \(\mathscr V\)-enriched category theory The bicategory \(\mathscr V\Prof\) has objects small \(\mathscr V\)-categories, \(1\)-cells profunctors, and \(2\)-cells \(\mathscr V\)-natural transformations. Composition is by coend, \[ (\psi\otimes\phi)(c,a)=\int^b \phi(b,a)\otimes\psi(c,b), \] and the identity on \(\mathscr A\) is the hom-profunctor \(\mathscr A(-,-)\). For a \(\mathscr V\)-functor \(F:\mathscr A\to\mathscr B\), the counit of the adjunction \(F_*\dashv F^*\) has components \[ \eps^F_{b,b'}:\int^a \mathscr B(Fa,b')\otimes\mathscr B(b,Fa)\longrightarrow \mathscr B(b,b'). \] Cauchy density is exactly the requirement that these components be isomorphisms in \(\mathscr V\) for every pair \(b,b'\in\mathscr B\) [2507.07869]. Several equivalent characterizations are available. For \(F:\mathscr A\to\mathscr B\) between small \(\mathscr V\)-categories, the following are equivalent: \(F\) is a lax epimorphism in \(\mathscr V\Cat\); for each small \(\mathscr C\), the restriction \([F,\mathscr C]:[\mathscr B,\mathscr C]\to[\mathscr A,\mathscr C]\) is fully faithful; \(\Lan_F\mathscr B(b,F-)\cong \mathscr B(b,-)\) for each \(b\in\mathscr B\); \(F\) is Cauchy dense; \(F\) is absolutely dense; and \(F\) is absolutely codense [2507.07869]. The theory emphasizes that Cauchy density is strictly stronger than “(ordinary) density” \(\Lan_FF\cong1\), but is self-dual and closed under composition. Dually, \(F\) is Cauchy dense if and only if \(F^{op}:\mathscr A^{op}\to\mathscr B^{op}\) is. When \(\mathscr V=\mathbb R^+\) in the sense of Lawvere metric spaces, one recovers topological density: a short map \(f:A\to B\) of classical metric spaces is Cauchy dense if and only if its image \(f(A)\) is topologically dense in \(B\) [2507.07869]. ## 3. Universal property of Cauchy completion The Cauchy completion \(\overline{\mathscr A}\subset[\mathscr A^{op},\mathscr V]\) consists of those presheaves which are “absolute weights” or equivalently “small-projective.” Within this framework, the Yoneda embedding \(z:\mathscr A\hookrightarrow\overline{\mathscr A}\) exhibits \(\overline{\mathscr A}\) as the largest \(\mathscr V\)-category into which \(\mathscr A\) admits a fully faithful Cauchy dense functor [2507.07869]. Concretely, whenever \(F:\mathscr A\to\mathscr B\) is fully faithful and Cauchy dense, there is an essentially unique fully faithful \[ N_F:\mathscr B\longrightarrow\overline{\mathscr A} \] making \[ \mathscr A\xrightarrow{\,F\,}\mathscr B\xrightarrow{\,N_F\,}\overline{\mathscr A} \quad\cong\quad \mathscr A\xrightarrow{\,z\,}\overline{\mathscr A} \] commute up to isomorphism. Moreover, any full subcategory of \(\overline{\mathscr A}\) containing the representables remains Cauchy complete, and the inclusion \(z\) is Cauchy dense [2507.07869]. A related lemma isolates a sufficient hypothesis for representable-on-the-left profunctors to lie in the completion: if \(F:\mathscr A\to\mathscr B\) is split-full and Cauchy dense, then each \(\mathscr B(F-,b):\mathscr A^{op}\to\mathscr V\) lies in \(\overline{\mathscr A}\). The result gives a direct route from Cauchy density to the small-projective structure that defines Cauchy completion. ## 4. Morita-type equivalence and special cases The paper establishes a Morita-type criterion for fully faithful Cauchy dense functors. If \(F:\mathscr A\to\mathscr B\) is a functor between small \(\mathscr V\)-categories, then \[ F\text{ is fully faithful and Cauchy dense} \quad\Longleftrightarrow\quad \text{for every Cauchy complete }\mathscr C,\;[F,\mathscr C]:[\mathscr B,\mathscr C]\to[\mathscr A,\mathscr C]\text{ is an equivalence.} \] The forward direction factors through the Yoneda embeddings and the induced functor \(\overline{\mathscr B}\to\overline{\mathscr A}\); the reverse direction takes \(\mathscr C=\mathscr V^{op}\), which is Cauchy complete, and deduces both full faithfulness and Cauchy density by examining the induced adjunction between presheaf categories and restricting to the small-projective subcategories [2507.07869]. This perspective supports a Morita-equivalence statement: two small \(\mathscr V\)-categories \(\mathscr A\) and \(\mathscr B\) are Morita equivalent, \(\overline{\mathscr A}\simeq\overline{\mathscr B}\), if and only if they can be connected by a zig-zag of fully faithful Cauchy dense functors [2507.07869]. A plausible implication is that Cauchy density functions as the exact enriched analogue of “passing to the same Cauchy completion.” The examples and special cases are explicit. If \(\mathscr V_0\) is a preorder, then \(F\) is Cauchy dense if and only if the single-object counits \(\eps_{b,b}:\int^a\mathscr B(Fa,b)\otimes\mathscr B(b,Fa)\to\mathscr B(b,b)\) are isomorphisms for all \(b\); in that case only the image of \(F\) on objects matters. A \(\mathsf 2\)-functor \(f:A\to B\) of preorders is Cauchy dense if and only if \(f\) is essentially surjective. Viewed as ordinary categories, Cauchy density requires each interval \(\{\,a\mid b\le fa\le b'\,\}\) be connected whenever \(b\le b'\). For monoid homomorphisms \(f:A\to B\), Cauchy density is characterized by bijectivity of the canonical map \[ \alpha:(B\times B)/\!\sim_f \xrightarrow{\;\cong\;} B, \] where \(\sim_f\) is generated by \((bf(a),b')\sim_f(b,f(a)b')\); if \(A\) is a group, this is equivalent to surjectivity of \(f\). A functor between discrete categories is Cauchy dense if and only if it is bijective on objects, and a functor whose domain is a groupoid is Cauchy dense exactly when it is equivalent in \(\Cat\) to a disjoint union of surjective group homomorphisms. For ordinary Cauchy dense functors, one also has a bijection on connected components, \(\pi_0(\mathscr A)\cong\pi_0(\mathscr B)\) [2507.07869]. ## 5. The Cauchy probability density and a Newton-inspired derivation For real \(x\), location parameter \(x_0\in\mathbb R\), and scale \(\gamma>0\), the Cauchy probability density function is \[ f(x;x_0,\gamma)=\frac{1}{\pi\,\gamma\,[1+((x-x_0)/\gamma)^2]}. \] The cited work derives this density from a one-dimensional process inspired by the geometrical interpretation of Newton’s method [2407.09583]. The construction begins with a quadratic \[ p_2(x)=a(x-x_0)^2+a s^2, \] so that its vertex abscissa is \(x_0\) and \(s^2=(4ac-b^2)/(4a^2)\). Newton’s tangent-intersection step produces the map \[ X_{k+1}=g(X_k)=X_k-\frac{p_2(X_k)}{p_2'(X_k)} = x_0+\frac{(X_k-x_0)^2+s^2}{2(X_k-x_0)}. \] Iterating this map \(N\) times with arbitrary \(X_1\in\mathbb R\setminus\{x_0\}\) yields a sample \(\{X_k\}\) whose empirical histogram, for large \(N\), converges to a smooth limit density [2407.09583]. The derivation proceeds by change of variables and a functional equation. Denoting by \(f\) the limiting density of \(X\), and using the two monotonic branches of \(g^{-1}\), one has \[ P(u0\) cancels under rescaling \(s\), one obtains the canonical Cauchy form with \(\gamma=s\) [2407.09583]. The geometric parameters of the quadratic transfer directly to the distribution. Median and mode coincide at \(x_0=-b/(2a)\), the parabola’s vertex abscissa, and the scale is \(\gamma=s=\sqrt{4ac-b^2}/(2a)\). In the simplest case \(p_2(x)=x^2+1\), one has \(x_0=0\), \(s=1\), hence the standard Cauchy \(C(0,1)\). The same source states that the Cauchy density has no finite mean or variance, heavy tails \(\sim 1/x^2\), and fails the law of large numbers [2407.09583]. ## 6. Transformations, randomized families, and information geometry The Newton-inspired iteration also yields an explicit transformed distribution for distances between successive intersections. Defining \[ R=|X_{k+1}-X_k|, \] one obtains \[ R=h(X)=\frac{(X-x_0)^2+s^2}{2|X-x_0|},\qquad X\sim \mathrm{Cauchy}(x_0,s). \] The transformation-of-variables formula gives \[ f_R(r)=\sum_{i=1}^4 \frac{f_X(x_i)}{|h'(x_i)|}, \] where \(\{x_i\}\) are the four real solutions of \(h(x)=r\) in the monotonic intervals around \(x_0\pm s\). Carrying out the algebra yields, for \(r>s\), \[ f_R(r)=\frac{1}{\pi s r}\left[\frac{L_1^2+s^2}{L_1^2-s^2}+\frac{L_2^2+s^2}{L_2^2-s^2}\right], \] with \[ L_1=r+\sqrt{r^2-s^2},\qquad L_2=r-\sqrt{r^2-s^2}. \] The same paper gives an algorithm for generating uniform randoms via the Cauchy process: initialize \(y_1=\tan((u_1-0.5)\pi)\) from a seed \(u_1\sim \mathrm{Uniform}(0,1)\); iterate \(y_{k+1}=\tfrac12[y_k-1/y_k]\); and convert back via \(u_{k+1}=\tfrac12+\tfrac1\pi\arctan(y_{k+1})\). It follows from \(\arctan(\mathrm{Cauchy})\mapsto \mathrm{Uniform}\) that \(\{u_k\}\to \mathrm{i.i.d.}\,\mathrm{Uniform}(0,1)\). With sample size \(N=10^6\) and significance \(\alpha=0.01\), the reported goodness-of-fit values are Kolmogorov–Smirnov \(p\approx0.98\), Anderson–Darling \(p\approx0.80\), Watson \(p\approx0.99\), and Cramér–von Mises \(p\approx0.93\); since all p-values \(\gg \alpha\), the null \(H_0\) is accepted [2407.09583]. A distinct construction studies the Cauchy density in the Dirichlet problem for Laplace’s equation. On the half-plane, \[ u_{xx}(x,y)+u_{yy}(x,y)=0,\qquad x>0,\; y\in\mathbb R,\qquad u(0,y)=f(y), \] the Poisson-kernel solution is \[ u(x,y)=\int_{\mathbb R}\Phi(x,y-\xi)\,f(\xi)\,d\xi, \qquad \Phi(x,y)=\frac1\pi\frac{x}{x^2+y^2}. \] If \(f\) is a probability density on \(\mathbb R\), one may interpret \[ p(\xi;x,y)=\frac{\Phi(x,y-\xi)\,f(\xi)}{u(x,y)} \] as a two-parameter family of randomized probability densities. Writing \[ h(x,y-\xi)=\frac{x}{x^2+(y-\xi)^2}, \] the paper records derivative identities for \(h\), including harmonicity \(\partial_x^2 h+\partial_y^2 h=0\), first- and second-order formulas, and third-order identities. These identities make possible explicit evaluation of the Fisher information matrix and the structure tensor [1703.06782]. For the randomized Cauchy family, with \(\ell(\xi;x,y)=\ln p(\xi;x,y)\), the Fisher information matrix \[ g_{ij}(x,y)=\int_{\mathbb R}\partial_i\ell\,\partial_j\ell\,p\,d\xi \] has the closed form \[ g_{11}=(\partial_x\ln u)^2+\frac{2}{x^2},\qquad g_{12}=\partial_x\partial_y\ln u,\qquad g_{22}=(\partial_y\ln u)^2+\frac{2}{x^2}. \] In particular the metric is manifestly positive definite for all \(x>0\), with \[ \det g=\frac{2}{x^2}\left((\ln u)_x^2+(\ln u)_y^2+\frac{2}{x^2}\right)>0. \] The same source gives formulas for the third-order structure tensor \(T_{ijk}\) and states that positive-definiteness of \(g_{ij}\) yields a family of inequalities on \(u(x,y)\), satisfied by every nonnegative solution of the Dirichlet problem for Laplace’s equation [1703.06782]. ## 7. Conceptual landscape In enriched category theory, Cauchy density organizes the passage from a small \(\mathscr V\)-category to its Cauchy completion, identifies the largest target admitting a fully faithful Cauchy dense embedding, and characterizes Morita equivalence by zig-zags of such functors [2507.07869]. The notion is self-dual, stronger than ordinary density, and stable under composition. Its special cases recover topological density in metric enrichment and essentially surjective or surjective behavior in preorders, groups, and discrete settings. In probability and analysis, the Cauchy density is the canonical heavy-tailed density \[ \frac{1}{\pi\,\gamma\,[1+((x-x_0)/\gamma)^2]}, \] arising in the cited works from a Newton-tangent-intersection process and from Poisson-kernel randomization of harmonic functions [2407.09583]. In that literature it is associated with explicit transformation formulas, a uniform-generation procedure via the \(\arctan\) mapping, and a statistical-manifold structure whose Fisher information matrix can be written in closed form [1703.06782]. This suggests that “Cauchy density” is best understood not as a single technical object but as a family of mathematically precise notions whose common vocabulary derives from Cauchy completion, Cauchy-type kernels, and limiting constructions. Source: https://www.emergentmind.com/topics/cauchy-density