-
Evaluation of Posits for Spectral Analysis Using a Software-Defined Dataflow Architecture
Authors:
Sameer Deshmukh,
Daniel Khankin,
William Killian,
John Gustafson,
Elad Raz
Abstract:
Spectral analysis plays an important role in detection of damage in structures and deep learning. The choice of a floating-point format plays a crucial role in determining the accuracy and performance of spectral analysis. The IEEE Std 754\textsuperscript{TM} floating-point format (IEEE~754 for short) is supported by most major hardware vendors for ``normal'' floats. However, it has several limita…
▽ More
Spectral analysis plays an important role in detection of damage in structures and deep learning. The choice of a floating-point format plays a crucial role in determining the accuracy and performance of spectral analysis. The IEEE Std 754\textsuperscript{TM} floating-point format (IEEE~754 for short) is supported by most major hardware vendors for ``normal'' floats. However, it has several limitations. Previous work has attempted to evaluate posit format with respect to accuracy and performance. The accuracy of the posit has been established over IEEE~754 for a variety of applications. For example, our analysis of the Fast Fourier Transform shows 2x better accuracy when using a 32-bit posit vs. a 32-bit IEEE754 format. For spectral analysis, 32-bit posits are substantially more accurate than 32-bit IEEE~754 floats. Although posit has shown better accuracy than IEEE~754, a fair evaluation of posit with IEEE~754 format using a real hardware implementation has been lacking so far. A software simulation of posit format on an x86 CPU is about $\mathbf{69.3\times}$ slower than native IEEE~754 hardware for normal floats for a Fast Fourier Transform (FFT) of $\mathbf{2^{28}}$ points. We propose the use of a software-defined dataflow architecture to evaluate performance and accuracy of posits in spectral analysis. Our dataflow architecture uses reconfigurable logical elements that express algorithms using only integer operations. Our architecture does not have an FPU, and we express both IEEE~754 and posit arithmetic using the same integer operations within the hardware. On our dataflow architecture, the posit format is only $\mathbf{1.8\times}$ slower than IEEE~754 for a Fast Fourier Transform (FFT) of $\mathbf{2^{28}\approx 268}$ million points. With this implementation, we empirically propose a new lower bound for the performance of posit compared to IEEE~754 format.
△ Less
Submitted 8 June, 2024;
originally announced June 2024.
-
On rich lenses in planar arrangements of circles and related problems
Authors:
Esther Ezra,
Orit E. Raz,
Micha Sharir,
Joshua Zahl
Abstract:
We show that the maximum number of pairwise non-overlap** $k$-rich lenses (lenses formed by at least $k$ circles) in an arrangement of $n$ circles in the plane is $O\left(\frac{n^{3/2}\log{(n/k^3)}}{k^{5/2}} + \frac{n}{k} \right)$, and the sum of the degrees of the lenses of such a family (where the degree of a lens is the number of circles that form it) is…
▽ More
We show that the maximum number of pairwise non-overlap** $k$-rich lenses (lenses formed by at least $k$ circles) in an arrangement of $n$ circles in the plane is $O\left(\frac{n^{3/2}\log{(n/k^3)}}{k^{5/2}} + \frac{n}{k} \right)$, and the sum of the degrees of the lenses of such a family (where the degree of a lens is the number of circles that form it) is $O\left(\frac{n^{3/2}\log{(n/k^3)}}{k^{3/2}} + n\right)$. Two independent proofs of these bounds are given, each interesting in its own right (so we believe). We then show that these bounds lead to the known bound of Agarwal et al. (JACM 2004) and Marcus and Tardos (JCTA 2006) on the number of point-circle incidences in the plane. Extensions to families of more general algebraic curves and some other related problems are also considered.
△ Less
Submitted 7 December, 2020;
originally announced December 2020.
-
Subspace arrangements, graph rigidity and derandomization through submodular optimization
Authors:
Orit E. Raz,
Avi Wigderson
Abstract:
This paper presents a deterministic, strongly polynomial time algorithm for computing the matrix rank for a class of symbolic matrices (whose entries are polynomials over a field). This class was introduced, in a different language, by Lovász [Lov] in his study of flats in matroids, and proved a duality theorem putting this problem in $NP \cap coNP$. As such, our result is another demonstration wh…
▽ More
This paper presents a deterministic, strongly polynomial time algorithm for computing the matrix rank for a class of symbolic matrices (whose entries are polynomials over a field). This class was introduced, in a different language, by Lovász [Lov] in his study of flats in matroids, and proved a duality theorem putting this problem in $NP \cap coNP$. As such, our result is another demonstration where ``good characterization'' in the sense of Edmonds leads to an efficient algorithm. In a different paper Lovász [Lov79] proved that all such symbolic rank problems have efficient probabilistic algorithms, namely are in $BPP$. As such, our algorithm may be interpreted as a derandomization result, in the long sequence special cases of the PIT (Polynomial Identity Testing) problem. Finally, Lovász and Yemini [LoYe] showed how the same problem generalizes the graph rigidity problem in two dimensions. As such, our algorithm may be seen as a generalization of the well-known deterministic algorithm for the latter problem.
There are two somewhat unusual technical features in this paper. The first is the translation of Lovász' flats problem into a symbolic rank one. The second is the use of submodular optimization for derandomization. We hope that the tools developed for both will be useful for related problems, in particular for better understanding of graph rigidity in higher dimensions.
△ Less
Submitted 27 January, 2019;
originally announced January 2019.
-
An o-minimal Szemerédi-Trotter theorem
Authors:
Saugata Basu,
Orit E. Raz
Abstract:
We prove an analog of the Szemerédi-Trotter theorem in the plane for definable curves and points in any o-minimal structure over an arbitrary real closed field $\mathrm{R}$. One new ingredient in the proof is an extension of the well known crossing number inequality for graphs to the case of embeddings in any o-minimal structure over an arbitrary real closed field.
We prove an analog of the Szemerédi-Trotter theorem in the plane for definable curves and points in any o-minimal structure over an arbitrary real closed field $\mathrm{R}$. One new ingredient in the proof is an extension of the well known crossing number inequality for graphs to the case of embeddings in any o-minimal structure over an arbitrary real closed field.
△ Less
Submitted 12 July, 2017; v1 submitted 22 November, 2016;
originally announced November 2016.
-
Configurations of lines in space and combinatorial rigidity
Authors:
Orit E. Raz
Abstract:
Let $L$ be a sequence $(\ell_1,\ell_2,\ldots,\ell_n)$ of $n$ lines in $\mathbb{C}^3$. We define the {\it intersection graph} $G_L=([n],E)$ of $L$, where $[n]:=\{1,\ldots, n\}$, and with $\{i,j\}\in E$ if and only if $i\neq j$ and the corresponding lines $\ell_i$ and $\ell_j$ intersect, or are parallel (or coincide). For a graph $G=([n],E)$, we say that a sequence $L$ is a {\it realization} of $G$…
▽ More
Let $L$ be a sequence $(\ell_1,\ell_2,\ldots,\ell_n)$ of $n$ lines in $\mathbb{C}^3$. We define the {\it intersection graph} $G_L=([n],E)$ of $L$, where $[n]:=\{1,\ldots, n\}$, and with $\{i,j\}\in E$ if and only if $i\neq j$ and the corresponding lines $\ell_i$ and $\ell_j$ intersect, or are parallel (or coincide). For a graph $G=([n],E)$, we say that a sequence $L$ is a {\it realization} of $G$ if $G\subset G_L$. One of the main results of this paper is to provide a combinatorial characterization of graphs $G=([n],E)$ that have the following property: For every {\it generic} realization $L$ of $G$ that consists of $n$ pairwise distinct lines, we have $G_L=K_n$, in which case the lines of $L$ are either all concurrent or all coplanar.
The general statements that we obtain about lines, apart from their independent interest, turns out to be closely related to the notion of graph rigidity. The connection is established due to the so-called Elekes--Sharir framework, which allows us to transform the problem into an incidence problem involving lines in three dimensions. By exploiting the geometry of contacts between lines in 3D, we can obtain alternative, simpler, and more precise characterizations of the rigidity of graphs.
△ Less
Submitted 14 July, 2016;
originally announced July 2016.
-
The Elekes-Szabó Theorem in four dimensions
Authors:
Orit E. Raz,
Micha Sharir,
Frank de Zeeuw
Abstract:
Let $F\in\mathbb{C}[x,y,s,t]$ be an irreducible constant-degree polynomial, and let $A,B,C,D\subset\mathbb{C}$ be finite sets of size $n$. We show that $F$ vanishes on at most $O(n^{8/3})$ points of the Cartesian product $A\times B\times C\times D$, unless $F$ has a special group-related form. A similar statement holds for $A,B,C,D$ of unequal sizes. This is a four-dimensional extension of our rec…
▽ More
Let $F\in\mathbb{C}[x,y,s,t]$ be an irreducible constant-degree polynomial, and let $A,B,C,D\subset\mathbb{C}$ be finite sets of size $n$. We show that $F$ vanishes on at most $O(n^{8/3})$ points of the Cartesian product $A\times B\times C\times D$, unless $F$ has a special group-related form. A similar statement holds for $A,B,C,D$ of unequal sizes. This is a four-dimensional extension of our recent improved analysis of the original Elekes-Szabó theorem in three dimensions. We give three applications: an expansion bound for three-variable real polynomials that do not have a special form, a bound on the number of coplanar quadruples on a space curve that is neither planar nor quartic, and a bound on the number of four-point circles on a plane curve that has degree at least five.
△ Less
Submitted 1 November, 2016; v1 submitted 13 July, 2016;
originally announced July 2016.
-
A note on distinct distances
Authors:
Orit E. Raz
Abstract:
We show that, for a constant-degree algebraic curve $γ$ in $\mathbb{R}^D$, every set of $n$ points on $γ$ spans at least $Ω(n^{4/3})$ distinct distances, unless $γ$ is an {\it algebraic helix} (see Definition 1.1). This improves the earlier bound $Ω(n^{5/4})$ of Charalambides [Discrete Comput. Geom. (2014)].
We also show that, for every set $P$ of $n$ points that lie on a $d$-dimensional constan…
▽ More
We show that, for a constant-degree algebraic curve $γ$ in $\mathbb{R}^D$, every set of $n$ points on $γ$ spans at least $Ω(n^{4/3})$ distinct distances, unless $γ$ is an {\it algebraic helix} (see Definition 1.1). This improves the earlier bound $Ω(n^{5/4})$ of Charalambides [Discrete Comput. Geom. (2014)].
We also show that, for every set $P$ of $n$ points that lie on a $d$-dimensional constant-degree algebraic variety $V$ in $\mathbb{R}^D$, there exists a subset $S\subset P$ of size at least $Ω(n^{\frac{4}{9+12(d-1)}})$, such that $S$ spans $\binom{|S|}{2}$ distinct distances. This improves the earlier bound of $Ω(n^{\frac{1}{3d}})$ of Conlon et al. [SIAM J. Discrete Math. (2015)].
Both results are consequences of a common technical tool, given in Lemma 2.7 below.
△ Less
Submitted 14 April, 2020; v1 submitted 29 February, 2016;
originally announced March 2016.
-
The number of unit-area triangles in the plane: Theme and variations
Authors:
Orit E. Raz,
Micha Sharir
Abstract:
We show that the number of unit-area triangles determined by a set $S$ of $n$ points in the plane is $O(n^{20/9})$, improving the earlier bound $O(n^{9/4})$ of Apfelbaum and Sharir [Discrete Comput. Geom., 2010]. We also consider two special cases of this problem: (i) We show, using a somewhat subtle construction, that if $S$ consists of points on three lines, the number of unit-area triangles tha…
▽ More
We show that the number of unit-area triangles determined by a set $S$ of $n$ points in the plane is $O(n^{20/9})$, improving the earlier bound $O(n^{9/4})$ of Apfelbaum and Sharir [Discrete Comput. Geom., 2010]. We also consider two special cases of this problem: (i) We show, using a somewhat subtle construction, that if $S$ consists of points on three lines, the number of unit-area triangles that $S$ spans can be $Ω(n^2)$, for any triple of lines (it is always $O(n^2)$ in this case). (ii) We show that if $S$ is a {\em convex grid} of the form $A\times B$, where $A$, $B$ are {\em convex} sets of $n^{1/2}$ real numbers each (i.e., the sequences of differences of consecutive elements of $A$ and of $B$ are both strictly increasing), then $S$ determines $O(n^{31/14})$ unit-area triangles.
△ Less
Submitted 11 April, 2015; v1 submitted 2 January, 2015;
originally announced January 2015.
-
Partial-Matching and Hausdorff RMS Distance Under Translation: Combinatorics and Algorithms
Authors:
Rinat Ben-Avraham,
Matthias Henze,
Rafel Jaume,
Balázs Keszegh,
Orit E. Raz,
Micha Sharir,
Igor Tubis
Abstract:
We consider the RMS distance (sum of squared distances between pairs of points) under translation between two point sets in the plane, in two different setups. In the partial-matching setup, each point in the smaller set is matched to a distinct point in the bigger set. Although the problem is not known to be polynomial, we establish several structural properties of the underlying subdivision of t…
▽ More
We consider the RMS distance (sum of squared distances between pairs of points) under translation between two point sets in the plane, in two different setups. In the partial-matching setup, each point in the smaller set is matched to a distinct point in the bigger set. Although the problem is not known to be polynomial, we establish several structural properties of the underlying subdivision of the plane and derive improved bounds on its complexity. These results lead to the best known algorithm for finding a translation for which the partial-matching RMS distance between the point sets is minimized. In addition, we show how to compute a local minimum of the partial-matching RMS distance under translation, in polynomial time. In the Hausdorff setup, each point is paired to its nearest neighbor in the other set. We develop algorithms for finding a local minimum of the Hausdorff RMS distance in nearly linear time on the line, and in nearly quadratic time in the plane. These improve substantially the worst-case behavior of the popular ICP heuristics for solving this problem.
△ Less
Submitted 26 November, 2014;
originally announced November 2014.
-
Polynomials vanishing on grids: The Elekes-Rónyai problem revisited
Authors:
Orit E. Raz,
Micha Sharir,
József Solymosi
Abstract:
In this paper we characterize real bivariate polynomials which have a small range over large Cartesian products. We show that for every constant-degree bivariate real polynomial $f$, either $|f(A,B)|=Ω(n^{4/3})$, for every pair of finite sets $A,B\subset{\mathbb R}$, with $|A|=|B|=n$ (where the constant of proportionality depends on ${\rm deg} f$), or else $f$ must be of one of the special forms…
▽ More
In this paper we characterize real bivariate polynomials which have a small range over large Cartesian products. We show that for every constant-degree bivariate real polynomial $f$, either $|f(A,B)|=Ω(n^{4/3})$, for every pair of finite sets $A,B\subset{\mathbb R}$, with $|A|=|B|=n$ (where the constant of proportionality depends on ${\rm deg} f$), or else $f$ must be of one of the special forms $f(u,v)=h(\varphi(u)+ψ(v))$, or $f(u,v)=h(\varphi(u)\cdotψ(v))$, for some univariate polynomials $\varphi,ψ,h$ over ${\mathbb R}$. This significantly improves a result of Elekes and Rónyai (2000).
Our results are cast in a more general form, in which we give an upper bound for the number of zeros of $z=f(x,y)$ on a triple Cartesian product $A\times B\times C$, when the sizes $|A|$, $|B|$, $|C|$ need not be the same; the upper bound is $O(n^{11/6})$ when $|A|=|B|=|C|=n$, where the constant of proportionality depends on ${\rm deg} f$, unless $f$ has one of the aforementioned special forms.
This result provides a unified tool for improving bounds in various Erd\H os-type problems in geometry and additive combinatorics. Several applications of our results to problems of these kinds are presented. For example, we show that the number of distinct distances between $n$ points lying on a constant-degree parametric algebraic curve which does not contain a line, in any dimension, is $Ω(n^{4/3})$, extending the result of Pach and de Zeeuw (2013) and improving the bound of Charalambides (2012), for the special case where the curve under consideration has a polynomial parameterization. We also derive improved lower bounds for several variants of the sum-product problem in additive combinatorics.
△ Less
Submitted 19 March, 2014; v1 submitted 29 January, 2014;
originally announced January 2014.
-
On the zone of the boundary of a convex body
Authors:
Orit Esther Raz
Abstract:
We consider an arrangement $\A$ of $n$ hyperplanes in $\R^d$ and the zone $\Z$ in $\A$ of the boundary of an arbitrary convex set in $\R^d$ in such an arrangement. We show that, whereas the combinatorial complexity of $\Z$ is known only to be $O<n^{d-1}\log n>$ \cite{APS}, the outer part of the zone has complexity $O<n^{d-1}>$ (without the logarithmic factor). Whether this bound also holds for the…
▽ More
We consider an arrangement $\A$ of $n$ hyperplanes in $\R^d$ and the zone $\Z$ in $\A$ of the boundary of an arbitrary convex set in $\R^d$ in such an arrangement. We show that, whereas the combinatorial complexity of $\Z$ is known only to be $O<n^{d-1}\log n>$ \cite{APS}, the outer part of the zone has complexity $O<n^{d-1}>$ (without the logarithmic factor). Whether this bound also holds for the complexity of the inner part of the zone is still an open question (even for $d=2$).
△ Less
Submitted 10 June, 2013;
originally announced June 2013.
