Konstantin (Kostya) Rybnikov

Department of Mathematical Sciences
Olney Hall 428K, North Campus
University of Massachusetts
Lowell, MA 01854

Curriculum Vitae in PDF (as of February 2008)

Research Interests:

Geometry of Numbers,

Discrete Geometry and Convexity,

Computational Geometry and Topology,

Graph Rigidity & Flexibility with a View towards Molecular Modeling,

Topological and Geometric Graph Theory,

Stochastic Geometry and Percolation Theory,

Statistics.

 


I am willing and interested in supervising student research and advanced directed learning at undergraduate, Masters, and doctoral levels in some of the above areas and in implementation of geometric and topological algorithms. Interested students email to krybniko@cs.uml.edu . You can also contact me if you are interested in doctoral program administered jointly by departments of Mathematical Sciences and Computer Science called D. Sc. in Computer Science - Computational Mathematics Option  (this might also be called Mathematics of Computing and Computational Statistics). While the paperwork should go thought the UML Graduate School, you can contact me by email with information about your background, interests, and expectations. Funding is possible through various sources.

I am also willing to do summer undergraduate research (similar to National Science Foundation's REU) with students living / working / studying in Eastern Massachusetts in the spring and summer.

This is not an regular RA/TA position ad -- please do not contact me if you are only looking for supplementing your income with a TA or RA contract, but your interests lie in other areas.

 


 

Selected Papers and Preprints

 

On Convexity of Hypersurfaces in the Hyperbolic Space (preprint)

(posted in February of 2008)

Abstract: In the Hyperbolic space  Hn (n >2) there are uncountably many topological types of convex hypersurfaces. When is a locally convex hypersurface in Hn  globally convex, that is, when does it bound a convex set? We prove that any locally convex proper embedding of an (n-1)-dimensional connected manifold is the boundary of a convex set whenever the complement of (n-1)-flats of the resulting hypersurface is connected.

PDF

 


 

Packing Density as a Function on the Voronoi Graph of Perfect Forms (preprint)

Abstract:  We observe that for small n the ball packing density does not have suboptimal  local maxima on the reduced Voronoi graph, whose vertices are the arithmetic equivalence classes of perfect quadratic forms and whose edges are the walls between the corresponding perfect domains.

PDF

 


 

An Efficient Local Approach to Convexity Testing of Piecewise-Linear Hypersurfaces (to appear in Computational Geometry: Theory and Applications)

(updated in January of 2008)

Abstract: We show that a closed connected bounded piecewise-linear hypersurface in the Euclidean space of dimension at least 3 is the boundary of a convex body if and only if every point in the interior of each (n-3)-face has a neighborhood that lies on the boundary of some convex body; no assumptions about the hypersurface topology are needed. We derive this criterion from our generalization of van Heijenoort's (1952) theorem on locally convex hypersurfaces in Euclidean spaces  to spherical spaces. We also give an easy-to-implement convexity testing algorithm, which is based on our criterion. For the surfaces in the 3-dimensional Euclidean space  the number of arithmetic operations used by the algorithm is at most linear in the number of vertices, while in general it is at most linear in the number of incidences between the (n-2)-faces and (n-3)-faces. When the dimension n is not fixed and only ring arithmetic (+,-,x) is allowed, the algorithm still remains polynomial. Our method works in more general situations than the convexity verification algorithms developed by Mehlhorn et al. (1996) and Devillers et al. (1998) -- for example, our method does not require the in- put surface to be orientable, nor it requires the input data to include normal vectors to the facets that are oriented ``in a coherent way". For the 3-dimensional space  the complexity of our algorithm is the same as that of previous algorithms; for higher dimensions there seems to be no clear winner, but our approach is the only one that easily handle inputs in which the facet normals are not known to be coherently oriented or are not given at all. Furthermore, our method can be extended to piecewise-polynomial surfaces of small degree.

PDF

 


 

Perfect Delaunay Polytopes and Perfect Quadratic Functions on Lattices (with R. M. Erdahl and A. Ordine)

 Contemporary Mathematics  452 (2008).

 
Abstract:  A polytope D whose vertices belong to a lattice of rank d is Delaunay if there is a circumscribing d-dimensional ellipsoid, E, with interior free of lattice points  so that the vertices of D lie on E. If in addition, the ellipsoid E  is uniquely determined by D, we call D perfect. That is, a perfect Delaunay polytope is a lattice polytope with a circumscribing empty ellipsoid E, where the surface of E both contains the vertices of D and is determined by them. We have been able to construct infinite sequences of perfect Delaunay polytopes, one perfect polytope in each successive dimension starting at some initial dimension; we have been able to construct an infinite number of such infinite sequences. Perfect Delaunay polytopes play an important role in the theory of Delaunay polytopes, and in Voronoi's theory of lattice types.

PDF

 


 

A New Algorithm in Geometry of Numbers (with M. Dutour)

In Proceedings of ISVD-07, pp. 182-189, IEEE International Symposium on Voronoi Diagrams in Science and Engineering, Pontypridd,

Wales, July 2007. Published by IEEE Publishing Services, Los Angeles, USA, 2007.

Abstract:  A lattice Delaunay polytope P is called perfect if its Delaunay sphere is the only ellipsoid circumscribed about P. We present a new algorithm for finding perfect Delaunay polytopes. Our method overcomes the major shortcomings of the previously used method [Du05]. We have implemented and used our algorithm for finding perfect Delaunay polytopes in dimensions 6, 7, 8. Our findings lead to a new conjecture that sheds light on the structure of lattice Delaunay tilings.

 

PDF

 


 

Perfect Delaunay Polytopes in Low Dimensions (with M. Dutour, R. Erdahl)

INTEGERS: ELECTRONIC JOURNAL OF COMBINATORIAL NUMBER THEORY  7 (2007), #A39

http://www.westga.edu/~integers

 

Abstract: A lattice Delaunay polytope is perfect if its Delaunay sphere is its only circumscribed ellipsoid. A perfect Delaunay polytope naturally corresponds to a positive quadratic function on Zn that can be recovered uniquely from the data consisting of its minimum and all points of Zn where this minimum is achieved – a quadratic function with this uniqueness property is also called perfect. We develop a structural theory of perfect Delaunay polytopes and quadratic functions. We also describe all known perfect Delaunay polytopes in dimensions one through eight: our conjecture is that this list is complete.

 

 

PDF

 


 

Local Minimax Learning of Approximately Polynomial Functions (with L. Jones)

In Proceeding of Dagstuhl Seminar 06201 "Combinatorial and Algorithmic Foundations of Pattern and Association Discovery", held in May of 2006 in the International Conference and Research Center (IBFI), Schloss Dagstuhl. Available on-line at http://drops.dagstuhl.de/opus/volltexte/2007/891/

 

PDF

 


 

On Conway's Thrackle Conjecture (with W. Li, K. Daniels)

A shorter version, under the title ``A Study of Conway's Thrackle Conjecture", appeared in Proceedings of 18th Canadian Conference on Computational Geometry, August 14-16, 2006, Queen's University, Kingson, Canada.

Abstract: A thrackle is a drawing of a simple graph on the plane, where each edge is drawn as a smooth arc with distinct end-points, and every two arcs have exactly one common point, at which they have distinct tangents. Conway, who coined the term thrackle, conjectured that there is no thrackle with more edges than vertices -- a question which is still unsolved. A full thrackle is one with n vertices and n edges, and it is called non-extensible if it cannot be a subthrackle of a counterexample to Conway's conjecture on n vertices. We define the notion of incidence type for a thrackle, which is the sequence of degrees of all vertices in increasing order. We introduce three reduction operations that can be applied to full subthrackles of thrackles. These reductions enable us to rule out the extensibility of many infinite series of incidence types of full thrackles. After defining the 1-2-3 set, we reduce Conway's conjecture to the problem of proving that thrackles from the 1-2-3 set are not extensible. Our result proves the hypothesis of Wehner, who predicted that a potential counterexample to Conway's conjecture would have certain graph-theoretic properties.

 

PDF

 

 


 

Estimation of Euler Characteristic from Point Data  (with D. Klain, K. Daniels, B. Jones, C. Neacsu)  

NSF report

Abstract:  Determination of the geometry and topology of a 3-dimensional body is an important problem appearing in areas such as computer-aided design, medical tomography, crystallography, and molecular biology.  This paper estimates the Euler characteristic of a body from a finite sampling of points.  Unlike other approaches the body is not reconstructed.  Techniques from integral geometry are used to numerically estimate the Euler characteristic.  It is recursively calculated with respect to dimension using projections of points onto hyperplanes as well as the inclusion-exclusion and homotopy-invariance properties of the Euler characteristic.  The strategy is shown to perform efficiently and robustly on test data that includes random point samples of a body as well as lattice points.  Test data consists of a two-dimensional MRI image and a collection of three-dimensional bodies with various types of holes.  Guidance on selecting parameter values is provided.

 

PDF

 

Software (updated 5  June 2007 07:06:22 -0400):

EulerChi release2

Release 2 notes:

A bug in the parser part of  main.ccp was fixed (the bug resulted in the program refusing correctly formatted test description files).
No changes to the algorithms have been made.


 

Supertopes

with Introduction to Perfect Ellipsoids


Abstract: A perfect (Delaunay) ellipsoid is an ellipsoid in n-dimensional Euclidean space that does not contain integral points in its interior, but is uniquely defined by integral points that lie on its surface. A perfect Delaunay polytope with respect to a positive quadratic form φ is a polytope with integral vertices that is circumscribed by a perfect Delaunay ellipsoid with an equation whose quadratic part is φ. This document has been corrected on January 15, 2005. Note that it represents the state of the area as of the end of 2002.

PDF


 

Cycle Criteria for Balance in Abelian Gain Graphs with Application to Piecewise-Linear Geometry (with T. Zaslavsky)

Discrete and Computational Geometry,  Volume 34,  No. 2,   (2005),  251--268

Abstract: A gain (a.k.a. voltage) graph is a triple (Γ,g,G), where Γ is a graph with arbitrary, but fixed orientations of edges, and g is a homomorphism from the free group on the edges of Γ to some group G. (Γ,g,G) is called balanced if all closed walks on Γ lie in the kernel of g. Thomas Zaslavsky and I derived a computationally efficient test, formulated in terms of the binary cycle space of (Γ,g,G), that can be used to detect if (Γ,g,G) is balanced for some choices of G, in particular when G is Abelian. Consider a gain graph with Abelian gain group having no odd torsion. If there is a basis of the graph’s binary cycle space, each of whose members can be lifted to a closed walk whose gain is the identity, then the gain graph is balanced, provided that the graph is finite or the group has no non-trivial infinitely 2-divisible elements.  In discrete geometry, there is a number of local-to-global results dealing with tilings of spaces of constant curvature and PL-realizations of manifolds in Rn. Several examples are: Coxeter classification of groups generated by reflections; Venkov-McMullen characterization of parallelohedra; A. D. Alexandrov's Tiling Theorem and Dolbilin's Extension Theorem, generalizing it; Voronoï's construction of generatrix; Maxwell correspondence, etc. Our work on gain graphs allows for topological generalizations of the "lifting part" of Voronoï's generatrix theorem and some other classical results in discrete geometry.

 

PDF


Cycle and Circle Tests of Balance in Gain Graphs: Forbidden Minors and Their Groups  (with T. Zaslavsky),

Journal of Graph Theory,  Volume 51,  No. 1, (2006), 1--21 ; published on line August 8, 2005

Abstract: We examine two criteria for balance of a gain graph, one based on binary cycles and one on circles. The graphs for which each criterion is valid depend on the set of allowed gain groups. The binary cycle test is invalid, except for forests, if any possible gain group has an element of odd order. Assuming all groups are allowed, or all Abelian groups, or merely the cyclic group of order 3, we characterize, both constructively and by forbidden minors, the graphs for which the circle test is valid. It turns out that these three classes of groups have the same set of forbidden minors. The exact reason for the importance of the ternary cyclic group is not clear.

 

PDF


New Infinite Series of Perfect Quadratic Forms and Big Delaunay Simplexes in Zn (with R. Erdahl)

Proceedings of Steklov Mathematical Institute, Volume 239, No. 4,  (2002), 159--167

Abstract: Georges Voronoï (1908-09) introduced two importantre duction methods for positive quadratic forms: the reduction with perfect forms, and the reduction with L-type domains. A quadratic form is perfect if can be reconstructed from all representations of its arithmetic minimum. Two forms have the same L-type if Delaunay tilings of their lattices are affinely equivalent. Delaunay (1937-38) asked about possible relative volumes of lattice Delaunay simplexes. We construct an infinite series of Delaunay simplexes of relative volume n-3, the best known as of 2002. This series gives rise to an infinite series of perfect forms that we refer to as {τn}n>3. Form τn has remarkable properties, e.g., τ5 ~D5, τ6 ~ E6*, τ7 ~ φ157;  for all n the L-type domain of τn is adjacent to the L-type domain of the 2-nd perfect form Dn. Perfect form τn is a direct n-dimensional generalization of Korkine and Zolotareff's 3-rd perfect form φ25 in 5 variables. It turns out that τn is equivalent to Anzin's (1991) form hn.

PDF


 

Voronoï-Dickson Hypothesis on Perfect Forms and L-types (with R. Erdahl)

This compressed version appeared in the Peter Gruber Festshrift:
Rendiconti del Circolo Matematiko di Palermo,
Serie II, Tomo LII, part I (2002), 279--296
A longer version available as math.MG/0112098 on http://arXiv.org .

Abstract: Georges Voronoï (1908, 1909) introduced two important reduction methods for positive quadratic forms: the reduction with perfect forms, and the reduction with L-domains, often called domains of Delaunay type. The first method is important in studies of dense lattice packings of spheres. The second method provided the key tools for finding the least dense lattice coverings with equal spheres in lower dimensions. In his investigations, Voronoï heavily relied on that in dimensions less than 6 the partition of the cone of positive quadratic forms into L-types refines the partition of this cone into perfect domains. Voronoï conjectured implicitely and Dickson (1972) explicitely that the L-partition is always a refinement of the partition into perfect domains. This was proved for n<6 (Voronoï, Delaunay, Ryshkov, Baranovskii). We show that Voronoï-Dickson conjecture fails already in dimension 6.

The text contains a concise introduction to Voronoï's theories of perfect forms and L-types.  

PDF


 

The Loss of Tension in an Infinite Membrane with Holes Distributed According to a Poisson Law (with M. Menshikov and S. Volkov)

Advances in Applied Probability, Volume 34, (2002), 292--312

Abstract: What is the effect of punching holes at random in an infinite tensed membrane? When will the membrane still support tension? This problem was introduced by R. Connelly in connection with applications of rigidity theory to material sciences. The answer clearly depends on the shapes and the distribution of holes. We briefly outline a mathematical theory of tension based on graph rigidity theory and introduce a probabilistic model for this problem. We show that if the centers of the holes are distributed in R2 according to a Poisson law with density λ>0, and the shapes are i. i. d. and independent of the location of their centers, the tension is lost on all of R2 for any λ>0. After introducing a certain step-by-step dynamics for the loss of tension, we establish the existence of a non-random N=N(λ) such that N or less steps are a. s. enough for the loss of tension. Also, we prove that N(λ) >1 a. s. for sufficiently small λ>0. The processes described in the paper are somewhat related to bootstrap and rigidity percolation models.

PDF

 


 

During the summer of 2001 I directed a Research Experience for Undergraduates project funded by NSF:

Cornell REU 2001: Geometry of Numbers

 


 

Percolation of the Loss of Tension in an Infinite Triangular Lattice  (with R. Connelly and S. Volkov)

Journal of Statistical Physics, Volume 105, No. 1/2, (2001), 145--173

Abstract: We introduce a new class of bootstrap percolation models where the local rules are of a geometric nature as opposed to simple counts of standard bootstrap percolation. Our geometric bootstrap percolation comes from rigidity theory and convex geometry. We outline two percolation models: a Poisson model and a lattice model. Our Poisson model describe show defects---holes is one of the possible interpretations of these defects---imposed on a tensed membrane result in a redistribution or loss of tension in this membrane; the lattice model is motivated by applications of Hooke spring networks to problems in material sciences. An analysis of the Poisson model is given by Menshikov, Rybnikov, and Volkov (2002).  In the discrete set-up we consider regular and generic triangular lattices on the plane where each bond is removed with probability 1-p. The problem of the existence of tension on such lattice is solved by reducing it to a bootstrap percolation model where the set of local rules follows from the geometry of stresses. We show that both regular and perturbed lattices cannot support tension for any p<1. Moreover, the complete relaxation of tension -- as defined in Section 3 -- occurs in a finite time almost surely. Furthermore, we underline striking similarities in the properties of the Poisson and lattice models.

 

PDF

 


 

 

On Traces of d-stresses in the Skeletons of Lower Dimensions of Homology d-manifolds  (with R. Erdahl and S. Ryshkov)

European Journal of Combinatorics, Volume 22, No. 6, (2001), 801--820

Abstract: We show how a d-stress on a piecewise-linear realization of an oriented (non-simplicial, in general) d-manifold in Rd  naturally induces stresses of lower dimensions on this manifold, and discuss implications of this construction to the analysis of self-stresses in spatial frameworks. The mappings we construct are not linear, but polynomial. In the 1860-70s J. C. Maxwell described an interesting relationship between self-stresses in planar frameworks and vertical projections of polyhedral 2-dimensional surfaces from R3 . We offer a spatial analog of Maxwell's correspondence based on our polynomial mappings. By applying our main result we derive a class of 3-dimensional spider webs similar to the 2-dimensional spider webs described by Maxwell. We also conjecture an important property of our mappings that is based on the lower bound theorem ( g2(d+1) = dim Stress2 ≥ 0 ) for d-pseudomanifolds generically realized in Rd +1 proved by Fogelsanger.

This paper is probably the best source (on this web-site) on Maxwell correspondence.

Pre-publication version:  

PDF

PS

 


 

Stresses and Liftings of Cell-complexes

Discrete and Computational Geometry, Volume  21, (1999), 481--517

Abstract: This paper introduces a general notion of stress on cell-complexes immersed in Rd and reports on connections between stresses and liftings (generalization of C10-splines) of   d-dimensional cell-complexes immersed in Rd . New sufficient conditions for the existence of a sharp lifting for a "flat" piecewise-linear realization of a manifold are given. Our approach also gives some new results on the equivalence between spherical complexes and convex and star polytopes. As an application, two algorithms are given that determine whether a piecewise-linear realization of a d-manifold in Rd admits a lifting to Rd+1 , which satisfies given affine constraints. We also demonstrate connections between stresses and Voronoï-Dirichlet decompositions and show that any weighted Voronoï-Dirichlet decomposition, without non-compact cells, can be seen a weighted Delaunay decomposition and vice versa.

This paper is a good source on generalized Maxwell-Cremona correspondence

zipped PDF

gzipped PS

Corrigendum:

1.      Please note that Lemma 4.1 actually holds for the case when the graph is finite and the group G is Abelian without odd torsion. If the graph is infinite, the homomorphic image of the fundamental group of the graph should not be 2-divisible for Lemma 4.1 to hold true. Only torsion-free Abelian groups are used in the theorems of this paper.

2.      All results on splines (Section 8) hold not only for Crr-1-splines,but also for Ckr-1-splines, for any k<r (only the dimensions of spline spaces should be recalculated). The given proof goes through without modifications.


 

The Theory of Quality Translations with Applications to Tilings (with S. Ryshkov)

European  Journal of. Combinatorics  18, (1997), 431 – 444.

 

Abstract: In this paper, we consider a broad class of simply connected complexes that includes, for example, the face-to-face tilings of En and Sn . Along with a complex (or tiling) we consider a set Q of qualities that individually can be assigned to the various cells or tiles of the complex, and a group G which permutes the elements of Q. These qualities might be, for example, colors,  positive real numbers, or affine functions . If each ordered pair of adjacent cells is associated with a group element g from G, and a quality q of Q is assigned to any particular cell of the complex, qualities can be assigned to adjacent cells using the action of the group G on Q. In fact, since the complex is assumed to be simply connected (and therefore connected), a quality can be assigned to any cell of the complex by translating (in retrospect, ``transferring" would be a better word) qualities along paths of adjacent cells using the group action. Our main theorem gives sufficient conditions that the quality assigned to a cell in this manner is independent of the particular path used for the translation, and is therefore uniquely determined. As applications of this theorem we establish an n -dimensional generalization of the converse of Maxwell’s theorem on frameworks, obtain some theorems on coloring the cells of the complex, and generalize the main part of the famous theorem of Voronoi on primitive parallelohedra.

 

PDF

 


 

Generatrice: the Problems of Maxwell and Voronoï  (with S. Ryshkov)

 Doklady Mathematics, Volume 54, (1996), 614--617.

Abstract: The problem of determining whether a given tiling of the Euclidean space can be obtained as the projection of a convex surface has at least two origins -- Maxwell correspondence in Rigidity Theory and Voronoï's Generatrice in Geometry of Numbers and Mathematical Crystallography. We give a few necessary conditions for the plane and for the general dimension, including the theorem that for any (d-3)-simple tiling can be lifted to a convex surface. We also introduce the projective notion of a dual tiling that generalizes the notion of dual graph. It turns out that a convex tiling can be lifted to a convex surface if and only if a dual convex tiling exists. Unfortunately this work is available only in paper form. Email your requests.


Polyhedral Partitions and Stresses, Ph. D. Thesis, Department of Mathematics and Statistics, Queen's University, October 1999

(Degree granted in the spring of 2000)

Includes: generalized Maxwell-Cremona, partial Maxwell-Cremona correspondence for graph realized in 3-space, fast algorithm for lifting a (d-3)-simple tiling of Rd (equivalently, Schlegel diagram) to a convex hypersurface in Rd+1 , stochastic geometry models motivated by rigidity theory and statistical physics.

Corrigendum:

1.      Please note that Lemma 4.1 actually holds for the case when the graph is finite and the group G is Abelian without odd torsion. If the graph is infinite, the homomorphic image of the fundamental group of the graph should not be 2-divisible for Lemma 4.1 to hold true. Only torsion-free Abelian groups are used in the theorems of this paper.

2.      All results on splines (Section 8) hold not only for Crr-1-splines,but also for Ckr-1-splines, for any k<r (only the dimensions of spline spaces should be recalculated). The given proof goes through without modifications.

gzipped PDF

gzipped PS

 


 

Software (by K. Ramanathan and K. Rybnikov)

Work in progress: C++ Library for Analysis of Infinitesimal (i.e. First-Order) Rigidity of Structures in Euclidean Space.

Currently only linear-algebraic part is publicly available.

 FORMAMENTUM - release 2 (zip file)

FORMAMENTUM - release 2 (bzip2ed tar file)

Brief Description (updated 04 May 2007)

 In all experiments on (scalar) rigidity matrices of size  3|V| x |E| = 550 x 13600 or less  the rank was found correctly.  (Note that the bound of 36 x 106  previously stated here was a typo.) This code can be compiled with   g++,  any version from 2.96 and up. It was tested with g++ 2.96.X, 3.4.3 and 4.0.2 20050901 (prerelease: SUSE Linux).

For random (for any i , j in  V  the edge (ij) exists with probability p)  graphs on 350-550 vertices the average vertex degree of 5.5 or higher always resulted in a rigid graph.



 

Updated 19 February 2008