arXiv subjects

Steven J. Gortler

Publications and source records attributed to Steven J. Gortler.

44 records · Page 3Linked to original sources

arXiv2014

From Shading to Local Shape

We develop a framework for extracting a concise representation of the shape information available from diffuse shading in a small image patch. This produces a mid-level scene descriptor, comprised of local shape distributions that are inferred separately at every image patch across multiple scales. The framework is based on a quadratic representation of local shape that, in the absence of noise, has guarantees on recovering accurate local shape and lighting. And when noise is present, the inferred local shape distributions provide useful shape information without over-committing to any particular image explanation. These local shape distributions naturally encode the fact that some smooth diffuse regions are more informative than others, and they enable efficient and robust reconstruction of object-scale shape. Experimental results show that this approach to surface reconstruction compares well against the state-of-art on both synthetic images and captured photographs.

cs.CV↗

arXiv2013

On affine rigidity

We define the notion of affine rigidity of a hypergraph and prove a variety of fundamental results for this notion. First, we show that affine rigidity can be determined by the rank of a specific matrix which implies that affine rigidity is a generic property of the hypergraph.Then we prove that if a graph is is $(d+1)$-vertex-connected, then it must be "generically neighborhood affinely rigid" in $d$-dimensional space. This implies that if a graph is $(d+1)$-vertex-connected then any generic framework of its squared graph must be universally rigid. Our results, and affine rigidity more generally, have natural applications in point registration and localization, as well as connections to manifold learning.

cs.CG↗

arXiv2012

Measurement Isomorphism of Graphs

The d-measurement set of a graph is its set of possible squared edge lengths over all d-dimensional embeddings. In this note, we define a new notion of graph isomorphism called d-measurement isomorphism. Two graphs are d-measurement isomorphic if there is agreement in their d-measurement sets. A natural question to ask is "what can be said about two graphs that are d-measurement isomorphic?" In this note, we show that this property coincides with the 2-isomorphism property studied by Whitney.

math.MG↗

arXiv2012

A geometrical approach to computing free energy landscapes from short-ranged potentials

Particles interacting with short-ranged potentials have attracted increasing interest, partly for their ability to model mesoscale systems such as colloids interacting via DNA or depletion. We consider the free energy landscape of such systems as the range of the potential goes to zero. In this limit, the landscape is entirely defined by geometrical manifolds, plus a single control parameter. These manifolds are fundamental objects that do not depend on the details of the interaction potential, and provide the starting point from which any quantity characterizing the system -- equilibrium or non-equilibrium -- can be computed for arbitrary potentials. To consider dynamical quantities we compute the asymptotic limit of the Fokker-Planck equation, and show that it becomes restricted to the low-dimensional manifolds connected by "sticky" boundary conditions. To illustrate our theory, we compute the low-dimensional manifolds for n<=8 identical particles, providing a complete description of the lowest-energy parts of the landscape including floppy modes with up to 2 internal degrees of freedom. The results can be directly tested on colloidal clusters. This limit is a novel approach for understanding energy landscapes, and our hope is that it can also provide insight into finite-range potentials.

math-ph↗

arXiv2012

Duals of Orphan-Free Anisotropic Voronoi Diagrams are Triangulations

We show that, under mild conditions on the underlying metric, duals of appropriately defined anisotropic Voronoi diagrams are embedded triangulations. Furthermore, they always triangulate the convex hull of the vertices, and have other properties that parallel those of ordinary Delaunay triangulations. These results apply to the duals of anisotropic Voronoi diagrams of any set of vertices, so long as the diagram is orphan-free.

cs.CG↗

arXiv2011

Orphan-Free Anisotropic Voronoi Diagrams

We describe conditions under which an appropriately-defined anisotropic Voronoi diagram of a set of sites in Euclidean space is guaranteed to be composed of connected cells in any number of dimensions. These conditions are natural for problems in optimization and approximation, and algorithms already exist to produce sets of sites that satisfy them.

cs.CG↗

arXiv2010

Characterizing Generic Global Rigidity

A d-dimensional framework is a graph and a map from its vertices to E^d. Such a framework is globally rigid if it is the only framework in E^d with the same graph and edge lengths, up to rigid motions. For which underlying graphs is a generic framework globally rigid? We answer this question by proving a conjecture by Connelly, that his sufficient condition is also necessary: a generic framework is globally rigid if and only if it has a stress matrix with kernel of dimension d+1, the minimum possible. An alternate version of the condition comes from considering the geometry of the length-squared mapping l: the graph is generically locally rigid iff the rank of l is maximal, and it is generically globally rigid iff the rank of the Gauss map on the image of l is maximal. We also show that this condition is efficiently checkable with a randomized algorithm, and prove that if a graph is not generically globally rigid then it is flexible one dimension higher.

math.MG↗

arXiv2000

Quantum versus Classical Learnability

We consider quantum versions of two well-studied classical learning models: Angluin's model of exact learning from membership queries and Valiant's Probably Approximately Correct (PAC) model of learning from random examples. We give positive and negative results for quantum versus classical learnability. For each of the two learning models described above, we show that any concept class is information-theoretically learnable from polynomially many quantum examples if and only if it is information-theoretically learnable from polynomially many classical examples. In contrast to this information-theoretic equivalence betwen quantum and classical learnability, though, we observe that a separation does exist between efficient quantum and classical learnability. For both the model of exact learning from membership queries and the PAC model, we show that under a widely held computational hardness assumption for classical computation (the intractability of factoring), there is a concept class which is polynomial-time learnable in the quantum version but not in the classical version of the model.

quant-ph↗