Computational Geometry

New submissions

• New submissions
• Cross-lists

New submissions for Fri, 24 Jan 20

[1]  arXiv:2001.08419 [pdf, other]

Title: Arrangements of Approaching Pseudo-Lines

Authors: Stefan Felsner, Alexander Pilz, Patrick Schnider

Subjects: Computational Geometry (cs.CG)

We consider arrangements of $n$ pseudo-lines in the Euclidean plane where each pseudo-line $\ell_i$ is represented by a bi-infinite connected $x$-monotone curve $f_i(x)$, $x \in \mathbb{R}$, s.t.\ for any two pseudo-lines $\ell_i$ and $\ell_j$ with $i < j$, the function $x \mapsto f_j(x) - f_i(x)$ is monotonically decreasing and surjective (i.e., the pseudo-lines approach each other until they cross, and then move away from each other). We show that such \emph{arrangements of approaching pseudo-lines}, under some aspects, behave similar to arrangements of lines, while for other aspects, they share the freedom of general pseudo-line arrangements.
For the former, we prove: 1. There are arrangements of pseudo-lines that are not realizable with approaching pseudo-lines. 2. Every arrangement of approaching pseudo-lines has a dual generalized configuration of points with an underlying arrangement of approaching pseudo-lines.
For the latter, we show: 1. There are $2^{\Theta(n^2)}$ isomorphism classes of arrangements of approaching pseudo-lines (while there are only $2^{\Theta(n \log n)}$ isomorphism classes of line arrangements). 2. It can be decided in polynomial time whether an allowable sequence is realizable by an arrangement of approaching pseudo-lines.
Furthermore, arrangements of approaching pseudo-lines can be transformed into each other by flipping triangular cells, i.e., they have a connected flip graph, and every bichromatic arrangement of this type contains a bichromatic triangular cell.

Cross-lists for Fri, 24 Jan 20

[2]  arXiv:2001.08540 (cross-list from math.OC) [pdf, other]

Title: Stochastic Item Descent Method for Large Scale Equal Circle Packing Problem

Authors: Kun He, Min Zhang, Jianrong Zhou, Yan Jin, Chu-min Li

Comments: 7 pages, 2 figures, 3 tables

Subjects: Optimization and Control (math.OC); Computational Geometry (cs.CG); Machine Learning (cs.LG); Numerical Analysis (math.NA)

Stochastic gradient descent (SGD) is a powerful method for large-scale optimization problems in the area of machine learning, especially for a finite-sum formulation with numerous variables. In recent years, mini-batch SGD gains great success and has become a standard technique for training deep neural networks fed with big amount of data. Inspired by its success in deep learning, we apply the idea of SGD with batch selection of samples to a classic optimization problem in decision version. Given $n$ unit circles, the equal circle packing problem (ECPP) asks whether there exist a feasible packing that could put all the circles inside a circular container without overlapping. Specifically, we propose a stochastic item descent method (SIDM) for ECPP in large scale, which randomly divides the unit circles into batches and runs Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm on the corresponding batch function iteratively to speedup the calculation. We also increase the batch size during the batch iterations to gain higher quality solution. Comparing to the current best packing algorithms, SIDM greatly speeds up the calculation of optimization process and guarantees the solution quality for large scale instances with up to 1500 circle items, while the baseline algorithms usually handle about 300 circle items. The results indicate the highly efficiency of SIDM for this classic optimization problem in large scale, and show potential for other large scale classic optimization problems in which gradient descent is used for optimization.

• New submissions
• Cross-lists