-
Statistical Inference on Grayscale Images via the Euler-Radon Transform
Authors:
Kun Meng,
Mattie Ji,
**yu Wang,
Kexin Ding,
Henry Kirveslahti,
Ani Eloyan,
Lorin Crawford
Abstract:
Tools from topological data analysis have been widely used to represent binary images in many scientific applications. Methods that aim to represent grayscale images (i.e., where pixel intensities instead take on continuous values) have been relatively underdeveloped. In this paper, we introduce the Euler-Radon transform, which generalizes the Euler characteristic transform to grayscale images by…
▽ More
Tools from topological data analysis have been widely used to represent binary images in many scientific applications. Methods that aim to represent grayscale images (i.e., where pixel intensities instead take on continuous values) have been relatively underdeveloped. In this paper, we introduce the Euler-Radon transform, which generalizes the Euler characteristic transform to grayscale images by using o-minimal structures and Euler integration over definable functions. Coupling the Karhunen-Loeve expansion with our proposed topological representation, we offer hypothesis-testing algorithms based on the chi-squared distribution for detecting significant differences between two groups of grayscale images. We illustrate our framework via extensive numerical experiments and simulations.
△ Less
Submitted 27 August, 2023;
originally announced August 2023.
-
Early Warning Signals of Social Instabilities in Twitter Data
Authors:
Vahid Shamsaddini,
Henry Kirveslahti,
Raphael Reinauer,
Wallyson Lemes de Oliveira,
Matteo Caorsi,
Etienne Voutaz
Abstract:
The goal of this project is to create and study novel techniques to identify early warning signals for socially disruptive events, like riots, wars, or revolutions using only publicly available data on social media. Such techniques need to be robust enough to work on real-time data: to achieve this goal we propose a topological approach together with more standard BERT models. Indeed, topology-bas…
▽ More
The goal of this project is to create and study novel techniques to identify early warning signals for socially disruptive events, like riots, wars, or revolutions using only publicly available data on social media. Such techniques need to be robust enough to work on real-time data: to achieve this goal we propose a topological approach together with more standard BERT models. Indeed, topology-based algorithms, being provably stable against deformations and noise, seem to work well in low-data regimes. The general idea is to build a binary classifier that predicts if a given tweet is related to a disruptive event or not. The results indicate that the persistent-gradient approach is stable and even more performant than deep-learning-based anomaly detection algorithms. We also benchmark the generalisability of the methodology against out-of-samples tasks, with very promising results.
△ Less
Submitted 3 March, 2023;
originally announced March 2023.
-
Representing Fields without Correspondences: the Lifted Euler Characteristic Transform
Authors:
Henry Kirveslahti,
Sayan Mukherjee
Abstract:
Topological transforms have been very useful in statistical analysis of shapes or surfaces without restrictions that the shapes are diffeomorphic and requiring the estimation of correspondence maps. In this paper we introduce two topological transforms that generalize from shapes to fields, $f:\mathbf{R}^3 \rightarrow \mathbf{R}$. Both transforms take a field and associate to each direction…
▽ More
Topological transforms have been very useful in statistical analysis of shapes or surfaces without restrictions that the shapes are diffeomorphic and requiring the estimation of correspondence maps. In this paper we introduce two topological transforms that generalize from shapes to fields, $f:\mathbf{R}^3 \rightarrow \mathbf{R}$. Both transforms take a field and associate to each direction $v\in S^{d-1}$ a summary obtained by scanning the field in the direction $v$. The transforms we introduce are of interest for both applications as well as their theoretical properties. The topological transforms for shapes are based on an Euler calculus on sets. A key insight in this paper is that via a lifting argument one can develop an Euler calculus on real valued functions from the standard Euler calculus on sets, this idea is at the heart of the two transforms we introduce. We prove the transforms are injective maps. We show for particular moduli spaces of functions we can upper bound the number of directions needed determine any particular function.
△ Less
Submitted 24 June, 2023; v1 submitted 8 November, 2021;
originally announced November 2021.
