Showing 1–8 of 8 results for author: Quercioli, N

Matching distance via the extended Pareto grid

Authors: Patrizio Frosini, Eloy Mósig García, Nicola Quercioli, Francesca Tombari

Abstract: One of the most animated themes of multidimensional persistence is the comparison between invariants. The matching distance between persistent Betti numbers functions (or rank invariants), is among the most studied metrics in this context, particularly in 2-parameter persistence. The main reason for this interest is that, in the 2-parameter case, the foliation method allows for an effective comput… ▽ More One of the most animated themes of multidimensional persistence is the comparison between invariants. The matching distance between persistent Betti numbers functions (or rank invariants), is among the most studied metrics in this context, particularly in 2-parameter persistence. The main reason for this interest is that, in the 2-parameter case, the foliation method allows for an effective computation of the matching distance, based on filtering the space along lines of positive slope. Our work provides a qualitative analysis, based on a construction called extended Pareto grid, of the filtering lines that actually contribute to the computation of the matching distance. Under certain genericity assumptions, we show that these lines must either be horizontal, vertical, of slope 1 or belong to a finite collection of special lines associated with discontinuity phenomena. △ Less

Submitted 6 July, 2024; v1 submitted 7 December, 2023; originally announced December 2023.

Comments: Theorem 5 is now a stronger result and Appendix B has been added

A topological model for partial equivariance in deep learning and data analysis

Authors: Lucia Ferrari, Patrizio Frosini, Nicola Quercioli, Francesca Tombari

Abstract: In this article, we propose a topological model to encode partial equivariance in neural networks. To this end, we introduce a class of operators, called P-GENEOs, that change data expressed by measurements, respecting the action of certain sets of transformations, in a non-expansive way. If the set of transformations acting is a group, then we obtain the so-called GENEOs. We then study the spaces… ▽ More In this article, we propose a topological model to encode partial equivariance in neural networks. To this end, we introduce a class of operators, called P-GENEOs, that change data expressed by measurements, respecting the action of certain sets of transformations, in a non-expansive way. If the set of transformations acting is a group, then we obtain the so-called GENEOs. We then study the spaces of measurements, whose domains are subject to the action of certain self-maps, and the space of P-GENEOs between these spaces. We define pseudo-metrics on them and show some properties of the resulting spaces. In particular, we show how such spaces have convenient approximation and convexity properties. △ Less

Submitted 25 August, 2023; originally announced August 2023.

Geometry of the matching distance for 2D filtering functions

Authors: Marc Ethier, Patrizio Frosini, Nicola Quercioli, Francesca Tombari

Abstract: In this paper we exploit the concept of extended Pareto grid to study the geometric properties of the matching distance for $\mathbb{R}^2$-valued regular functions defined on a Riemannian closed manifold. In particular, we prove that in this case the matching distance is realised either at special values or at values corresponding to vertical, horizontal or slope 1 lines. In this paper we exploit the concept of extended Pareto grid to study the geometric properties of the matching distance for $\mathbb{R}^2$-valued regular functions defined on a Riemannian closed manifold. In particular, we prove that in this case the matching distance is realised either at special values or at values corresponding to vertical, horizontal or slope 1 lines. △ Less

Submitted 7 December, 2023; v1 submitted 29 October, 2022; originally announced October 2022.

MSC Class: Primary 55N31; Secondary 57R19

On the geometric and Riemannian structure of the spaces of group equivariant non-expansive operators

Authors: Pasquale Cascarano, Patrizio Frosini, Nicola Quercioli, Amir Saki

Abstract: Group equivariant non-expansive operators have been recently proposed as basic components in topological data analysis and deep learning. In this paper we study some geometric properties of the spaces of group equivariant operators and show how a space $\mathcal{F}$ of group equivariant non-expansive operators can be endowed with the structure of a Riemannian manifold, so making available the use… ▽ More Group equivariant non-expansive operators have been recently proposed as basic components in topological data analysis and deep learning. In this paper we study some geometric properties of the spaces of group equivariant operators and show how a space $\mathcal{F}$ of group equivariant non-expansive operators can be endowed with the structure of a Riemannian manifold, so making available the use of gradient descent methods for the minimization of cost functions on $\mathcal{F}$. As an application of this approach, we also describe a procedure to select a finite set of representative group equivariant non-expansive operators in the considered manifold. △ Less

Submitted 31 December, 2023; v1 submitted 3 March, 2021; originally announced March 2021.

Comments: 21 pages, 1 figure. The introduction has been extended and a section on the group's action on the space of GENEOs has been added. Some minor fixes are made. The text has been simplified and made clearer

MSC Class: Primary: 55N31; Secondary: 58D20; 58D30; 62R40; 65D18; 68T09

Some new methods to build group equivariant non-expansive operators in TDA

Authors: Nicola Quercioli

Abstract: Group equivariant operators are playing a more and more relevant role in machine learning and topological data analysis. In this paper we present some new results concerning the construction of $G$-equivariant non-expansive operators (GENEOs) from a space $\varPhi$ of real-valued bounded continuous functions on a topological space $X$ to $\varPhi$ itself. The space $\varPhi$ represents our set of… ▽ More Group equivariant operators are playing a more and more relevant role in machine learning and topological data analysis. In this paper we present some new results concerning the construction of $G$-equivariant non-expansive operators (GENEOs) from a space $\varPhi$ of real-valued bounded continuous functions on a topological space $X$ to $\varPhi$ itself. The space $\varPhi$ represents our set of data, while $G$ is a subgroup of the group of all self-homeomorphisms of $X$, representing the invariance we are interested in. △ Less

Submitted 17 October, 2020; originally announced October 2020.

MSC Class: 55N31; 47H09; 54H15; 57S10; 68U05

On the finite representation of group equivariant operators via permutant measures

Authors: Giovanni Bocchi, Stefano Botteghi, Martina Brasini, Patrizio Frosini, Nicola Quercioli

Abstract: The study of $G$-equivariant operators is of great interest to explain and understand the architecture of neural networks. In this paper we show that each linear $G$-equivariant operator can be produced by a suitable permutant measure, provided that the group $G$ transitively acts on a finite signal domain $X$. This result makes available a new method to build linear $G$-equivariant operators in t… ▽ More The study of $G$-equivariant operators is of great interest to explain and understand the architecture of neural networks. In this paper we show that each linear $G$-equivariant operator can be produced by a suitable permutant measure, provided that the group $G$ transitively acts on a finite signal domain $X$. This result makes available a new method to build linear $G$-equivariant operators in the finite setting. △ Less

Submitted 9 March, 2022; v1 submitted 7 August, 2020; originally announced August 2020.

Comments: We have extended the introduction, providing more context about our approach, and the discussion. We have also added Section 4, showing possible advantages of the use of GENEOs built by permutant measures. The experiment described in Section 4 has been made by Giovanni Bocchi. For this reason, he has been added to the list of authors of the paper

MSC Class: Primary: 68T09 Secondary: 15B51; 20C35; 47B38; 55N31; 62R40; 68U05

Landscapes of data sets and functoriality of persistent homology

Authors: Wojciech Chacholski, Alessandro De Gregorio, Nicola Quercioli, Francesca Tombari

Abstract: The aim of this article is to describe a new perspective on functoriality of persistent homology and explain its intrinsic symmetry that is often overlooked. A data set for us is a finite collection of functions, called measurements, with a finite domain. Such a data set might contain internal symmetries which are effectively captured by the action of a set of the domain endomorphisms. Different c… ▽ More The aim of this article is to describe a new perspective on functoriality of persistent homology and explain its intrinsic symmetry that is often overlooked. A data set for us is a finite collection of functions, called measurements, with a finite domain. Such a data set might contain internal symmetries which are effectively captured by the action of a set of the domain endomorphisms. Different choices of the set of endomorphisms encode different symmetries of the data set. We describe various category structures on such enriched data sets and prove some of their properties such as decompositions and morphism formations. We also describe a data structure, based on coloured directed graphs, which is convenient to encode the mentioned enrichment. We show that persistent homology preserves only some aspects of these landscapes of enriched data sets however not all. In other words persistent homology is not a functor on the entire category of enriched data sets. Nevertheless we show that persistent homology is functorial locally. We use the concept of equivariant operators to capture some of the information missed by persistent homology. △ Less

Submitted 25 May, 2020; v1 submitted 14 February, 2020; originally announced February 2020.

Towards a topological-geometrical theory of group equivariant non-expansive operators for data analysis and machine learning

Authors: Mattia G. Bergomi, Patrizio Frosini, Daniela Giorgi, Nicola Quercioli

Abstract: The aim of this paper is to provide a general mathematical framework for group equivariance in the machine learning context. The framework builds on a synergy between persistent homology and the theory of group actions. We define group-equivariant non-expansive operators (GENEOs), which are maps between function spaces associated with groups of transformations. We study the topological and metric… ▽ More The aim of this paper is to provide a general mathematical framework for group equivariance in the machine learning context. The framework builds on a synergy between persistent homology and the theory of group actions. We define group-equivariant non-expansive operators (GENEOs), which are maps between function spaces associated with groups of transformations. We study the topological and metric properties of the space of GENEOs to evaluate their approximating power and set the basis for general strategies to initialise and compose operators. We begin by defining suitable pseudo-metrics for the function spaces, the equivariance groups, and the set of non-expansive operators. Basing on these pseudo-metrics, we prove that the space of GENEOs is compact and convex, under the assumption that the function spaces are compact and convex. These results provide fundamental guarantees in a machine learning perspective. We show examples on the MNIST and fashion-MNIST datasets. By considering isometry-equivariant non-expansive operators, we describe a simple strategy to select and sample operators, and show how the selected and sampled operators can be used to perform both classical metric learning and an effective initialisation of the kernels of a convolutional neural network. △ Less

Submitted 3 March, 2019; v1 submitted 31 December, 2018; originally announced December 2018.

Comments: Added references. Extended Section 7. Added 3 figures. Corrected typos. 42 pages, 7 figures

MSC Class: 55N35 (Primary); 47H09; 54H15; 57S10; 68U05; 65D18 (Secondary)