Recent work in neural abstract machines has proposed many useful techniques to learn sequences of applications of discrete but differentiable operators. These techniques allow us to model traditionally procedural problems using neural networks. In this work, we are interested in using neural networks to learn to perform logic reasoning. We propose a model that has access to differentiable opera...

Any nonpositively curved symmetric space admits a topological compactification, namely the Hadamard compactification. For rank 1 spaces, this topological compactification can be endowed with a differentiable structure such that the action of the isometry group is differentiable. Moreover, the restriction of the action on the boundary leads to a flat model for some geometry (conformal, CR or qua...

We find an equivalent condition for a real function to be Lebesgue equivalent to a twice differentiable function. For that purpose, we introduce the notion of a V BG 1 2 function, which plays an analogous rôle for the second order differentiability as the classical notion of a V BG∗ function for the first order differentiability. In fact, for a function f : [a, b] → R, being Lebesgue equivalent...

A classical result of Alexsandrov [1] asserts that convex functions in R are twice differentiable a.e., (see also [3], [8] for more modern treatments). It is well known that Sobolev functions u ∈ W , for p > n/2 are twice differentiable a.e.. The following weaker notion of convexity known as k-convexity was introduced by Trudinger and Wang [12, 13]. Let Ω ⊂ R be an open set and C(Ω) be the clas...