-
PCA for Point Processes
Authors:
Franck Picard,
Vincent Rivoirard,
Angelina Roche,
Victor Panaretos
Abstract:
We introduce a novel statistical framework for the analysis of replicated point processes that allows for the study of point pattern variability at a population level. By treating point process realizations as random measures, we adopt a functional analysis perspective and propose a form of functional Principal Component Analysis (fPCA) for point processes. The originality of our method is to base…
▽ More
We introduce a novel statistical framework for the analysis of replicated point processes that allows for the study of point pattern variability at a population level. By treating point process realizations as random measures, we adopt a functional analysis perspective and propose a form of functional Principal Component Analysis (fPCA) for point processes. The originality of our method is to base our analysis on the cumulative mass functions of the random measures which gives us a direct and interpretable analysis. Key theoretical contributions include establishing a Karhunen-Loève expansion for the random measures and a Mercer Theorem for covariance measures. We establish convergence in a strong sense, and introduce the concept of principal measures, which can be seen as latent processes governing the dynamics of the observed point patterns. We propose an easy-to-implement estimation strategy of eigenelements for which parametric rates are achieved. We fully characterize the solutions of our approach to Poisson and Hawkes processes and validate our methodology via simulations and diverse applications in seismology, single-cell biology and neurosiences, demonstrating its versatility and effectiveness. Our method is implemented in the pppca R-package.
△ Less
Submitted 30 April, 2024;
originally announced April 2024.
-
Adaptive nonparametric estimation in the functional linear model with functional output
Authors:
Gaëlle Chagny,
Anouar Meynaoui,
Angelina Roche
Abstract:
In this paper, we consider a functional linear regression model, where both the covariate and the response variable are functional random variables. We address the problem of optimal nonparametric estimation of the conditional expectation operator in this model. A collection of projection estimators over finite dimensional subspaces is first introduce. We provide a non-asymptotic bias-variance dec…
▽ More
In this paper, we consider a functional linear regression model, where both the covariate and the response variable are functional random variables. We address the problem of optimal nonparametric estimation of the conditional expectation operator in this model. A collection of projection estimators over finite dimensional subspaces is first introduce. We provide a non-asymptotic bias-variance decomposition for the Mean Square Prediction error in the case where these subspaces are generated by the (empirical) PCA functional basis. The automatic trade-off is realized thanks to a model selection device which selects the best projection dimensions: the penalized contrast estimator satisfies an oracle-type inequality and is thus optimal in an adaptive point of view. These upper-bounds allow us to derive convergence rates over ellipsoidal smoothness spaces. The rates are shown to be optimal in the minimax sense: they match with a lower bound of the minimax risk, which is also proved. Finally, we conduct a numerical study, over simulated data and over two real-data sets.
△ Less
Submitted 1 March, 2022;
originally announced March 2022.
-
Adaptive nonparametric estimation of a component density in a two-class mixture model
Authors:
Gaelle Chagny,
Antoine Channarond,
Van Ha Hoang,
Angelina Roche
Abstract:
A two-class mixture model, where the density of one of the components is known, is considered. We address the issue of the nonparametric adaptive estimation of the unknown probability density of the second component. We propose a randomly weighted kernel estimator with a fully data-driven bandwidth selection method, in the spirit of the Goldenshluger and Lepski method. An oracle-type inequality fo…
▽ More
A two-class mixture model, where the density of one of the components is known, is considered. We address the issue of the nonparametric adaptive estimation of the unknown probability density of the second component. We propose a randomly weighted kernel estimator with a fully data-driven bandwidth selection method, in the spirit of the Goldenshluger and Lepski method. An oracle-type inequality for the pointwise quadratic risk is derived as well as convergence rates over Holder smoothness classes. The theoretical results are illustrated by numerical simulations.
△ Less
Submitted 5 February, 2021; v1 submitted 30 July, 2020;
originally announced July 2020.
-
A note on dual modules and the transpose
Authors:
Thomas Madsen,
Alan Roche,
C. Ryan Vinroot
Abstract:
It is a classical result in matrix algebra that any square matrix over a field can be conjugated to its transpose by a symmetric matrix. For $F$ a non-Archimedean local field, Tupan used this to give an elementary proof that transpose inverse takes each irreducible smooth representation of ${\rm GL}_n(F)$ to its dual. We re-prove the matrix result and related observations using module-theoretic ar…
▽ More
It is a classical result in matrix algebra that any square matrix over a field can be conjugated to its transpose by a symmetric matrix. For $F$ a non-Archimedean local field, Tupan used this to give an elementary proof that transpose inverse takes each irreducible smooth representation of ${\rm GL}_n(F)$ to its dual. We re-prove the matrix result and related observations using module-theoretic arguments. In addition, we write down a generalization that applies to central simple algebras with an involution of the first kind. We use this generalization to extend Tupan's method of argument to ${\rm GL}_n(D)$ for $D$ a quaternion division algebra over $F$.
△ Less
Submitted 5 June, 2019;
originally announced June 2019.
-
Lasso in infinite dimension: application to variable selection in functional multivariate linear regression
Authors:
Angelina Roche
Abstract:
It is more and more frequently the case in applications that the data we observe come from one or more random variables taking values in an infinite dimensional space, e.g. curves. The need to have tools adapted to the nature of these data explains the growing interest in the field of functional data analysis. The model we study in this paper assumes a linear dependence between a quantity of inter…
▽ More
It is more and more frequently the case in applications that the data we observe come from one or more random variables taking values in an infinite dimensional space, e.g. curves. The need to have tools adapted to the nature of these data explains the growing interest in the field of functional data analysis. The model we study in this paper assumes a linear dependence between a quantity of interest and several covariates, at least one of which has an infinite dimension. To select the relevant covariates in this context, we investigate adaptations of the Lasso method. Two estimation methods are defined. The first one consists in the minimization of a Group-Lasso criterion on the multivariate functional space H. The second one minimizes the same criterion but on a finite dimensional subspaces of H whose dimension is chosen by a penalized least squares method. We prove oracle inequalities of sparsity in the case where the design is fixed or random. To compute the solutions of both criteria in practice, we propose a coordinate descent algorithm. A numerical study on simulated and real data illustrates the behavior of the estimators.
△ Less
Submitted 31 May, 2023; v1 submitted 29 March, 2019;
originally announced March 2019.
-
Local bandwidth selection for kernel density estimation in bifurcating Markov chain model
Authors:
S Valere Bitseki Penda,
Angelina Roche
Abstract:
We propose an adaptive estimator for the stationary distribution of a bifurcating Markov Chain on $\mathbb R^d$. Bifurcating Markov chains (BMC for short) are a class of stochastic processes indexed by regular binary trees. A kernel estimator is proposed whose bandwidth is selected by a method inspired by the works of Goldenshluger and Lepski [18]. Drawing inspiration from dimension jump methods f…
▽ More
We propose an adaptive estimator for the stationary distribution of a bifurcating Markov Chain on $\mathbb R^d$. Bifurcating Markov chains (BMC for short) are a class of stochastic processes indexed by regular binary trees. A kernel estimator is proposed whose bandwidth is selected by a method inspired by the works of Goldenshluger and Lepski [18]. Drawing inspiration from dimension jump methods for model selection, we also provide an algorithm to select the best constant in the penalty.
△ Less
Submitted 21 June, 2017;
originally announced June 2017.
-
Dualizing involutions for classical and similitude groups over local non-archimedean fields
Authors:
Alan Roche,
C. Ryan Vinroot
Abstract:
Building on ideas of Tupan, we give an elementary proof of a result of Mœglin, Vignéras and Waldspurger on the existence of automorphisms of many $p$-adic classical groups that take each irreducible smooth representations to its dual. Our proof also applies to the corresponding similitude groups.
Building on ideas of Tupan, we give an elementary proof of a result of Mœglin, Vignéras and Waldspurger on the existence of automorphisms of many $p$-adic classical groups that take each irreducible smooth representations to its dual. Our proof also applies to the corresponding similitude groups.
△ Less
Submitted 27 July, 2016;
originally announced July 2016.
-
A factorization result for classical and similitude groups
Authors:
Alan Roche,
C. Ryan Vinroot
Abstract:
For most classical and similitude groups, we show that each element can be written as a product of two transformations that a) preserve or almost preserve the underlying form and b) whose squares are certain scalar maps. This generalizes work of Wonenburger and Vinroot. As an application, we re-prove and slightly extend a well-known result of Mœglin, Vignéras and Waldspurger on the existence of au…
▽ More
For most classical and similitude groups, we show that each element can be written as a product of two transformations that a) preserve or almost preserve the underlying form and b) whose squares are certain scalar maps. This generalizes work of Wonenburger and Vinroot. As an application, we re-prove and slightly extend a well-known result of Mœglin, Vignéras and Waldspurger on the existence of automorphisms of $p$-adic classical groups that take each irreducible smooth representations to its dual.
△ Less
Submitted 22 July, 2016;
originally announced July 2016.
-
Local Optimization of Black-Box Function with High or Infinite-Dimensional Inputs
Authors:
Angelina Roche
Abstract:
An adaptation of Response Surface Methodology (RSM) when the covariate is of high or infinite dimensional is proposed, providing a tool for black-box optimization in this context. We combine dimension reduction techniques with classical multivariate Design of Experiments (DoE). We propose a method to generate experimental designs and extend usual properties (orthogonality, rotatability,...) of mul…
▽ More
An adaptation of Response Surface Methodology (RSM) when the covariate is of high or infinite dimensional is proposed, providing a tool for black-box optimization in this context. We combine dimension reduction techniques with classical multivariate Design of Experiments (DoE). We propose a method to generate experimental designs and extend usual properties (orthogonality, rotatability,...) of multivariate designs to general high or infinite dimensional contexts. Different dimension reduction basis are considered (including data-driven basis). The methodology is illustrated on simulated functional data and we discuss the choice of the different parameters, in particular the dimension of the approximation space. The method is finally applied to a problem of nuclear safety.
△ Less
Submitted 18 November, 2015; v1 submitted 9 June, 2015;
originally announced June 2015.
-
Non-asymptotic Adaptive Prediction in Functional Linear Models
Authors:
Elodie Brunel,
André Mas,
Angelina Roche
Abstract:
Functional linear regression has recently attracted considerable interest. Many works focus on asymptotic inference. In this paper we consider in a non asymptotic framework a simple estimation procedure based on functional Principal Regression. It revolves in the minimization of a least square contrast coupled with a classical projection on the space spanned by the m first empirical eigenvectors o…
▽ More
Functional linear regression has recently attracted considerable interest. Many works focus on asymptotic inference. In this paper we consider in a non asymptotic framework a simple estimation procedure based on functional Principal Regression. It revolves in the minimization of a least square contrast coupled with a classical projection on the space spanned by the m first empirical eigenvectors of the covariance operator of the functional sample. The novelty of our approach is to select automatically the crucial dimension m by minimization of a penalized least square contrast. Our method is based on model selection tools. Yet, since this kind of methods consists usually in projecting onto known non-random spaces, we need to adapt it to empirical eigenbasis made of data-dependent - hence random - vectors. The resulting estimator is fully adaptive and is shown to verify an oracle inequality for the risk associated to the prediction error and to attain optimal minimax rates of convergence over a certain class of ellipsoids. Our strategy of model selection is finally compared numerically with cross-validation.
△ Less
Submitted 14 January, 2013;
originally announced January 2013.
-
Signs, involutions and Jacquet modules
Authors:
Alan Roche,
Steven Spallone
Abstract:
Let $G$ be a connected reductive $p$-adic group and let $θ$ be an automorphism of $G$ of order at most two. Suppose $π$ is an irreducible smooth representation of $G$ that is taken to its dual by $θ$. The space $V$ of $π$ then carries a non-zero bilinear form $(\mspace{7mu},\mspace{6mu})$, unique up to scaling, with the invariance property $(π(g)v, π({}^θg)w) = (v,w)$, for $g \in G$ and…
▽ More
Let $G$ be a connected reductive $p$-adic group and let $θ$ be an automorphism of $G$ of order at most two. Suppose $π$ is an irreducible smooth representation of $G$ that is taken to its dual by $θ$. The space $V$ of $π$ then carries a non-zero bilinear form $(\mspace{7mu},\mspace{6mu})$, unique up to scaling, with the invariance property $(π(g)v, π({}^θg)w) = (v,w)$, for $g \in G$ and $v, w \in V$. The form is easily seen to be symmetric or skew-symmetric and we set $\varepsilon_θ(π) = \pm1$ accordingly. We use Cassleman's pairing (in commonly observed circumstances) to express $\varepsilon_θ(π)$ in terms of certain Jacquet modules of $π$ and thus, via the Langlands classification, reduce the problem of determining the sign to the case of tempered representations. For the transpose-inverse involution of the general linear group, we show that the associated signs are always one.
△ Less
Submitted 20 April, 2012;
originally announced April 2012.
-
Selection of a Model of Cerebral Activity for fMRI Group Data Analysis
Authors:
Merlin Keller,
Alexis Roche,
Marc Lavielle
Abstract:
This thesis is dedicated to the statistical analysis of multi-sub ject fMRI data, with the purpose of identifying bain structures involved in certain cognitive or sensori-motor tasks, in a reproducible way across sub jects. To overcome certain limitations of standard voxel-based testing methods, as implemented in the Statistical Parametric Mapping (SPM) software, we introduce a Bayesian model sele…
▽ More
This thesis is dedicated to the statistical analysis of multi-sub ject fMRI data, with the purpose of identifying bain structures involved in certain cognitive or sensori-motor tasks, in a reproducible way across sub jects. To overcome certain limitations of standard voxel-based testing methods, as implemented in the Statistical Parametric Mapping (SPM) software, we introduce a Bayesian model selection approach to this problem, meaning that the most probable model of cerebral activity given the data is selected from a pre-defined collection of possible models. Based on a parcellation of the brain volume into functionally homogeneous regions, each model corresponds to a partition of the regions into those involved in the task under study and those inactive. This allows to incorporate prior information, and avoids the dependence of the SPM-like approach on an arbitrary threshold, called the cluster- forming threshold, to define active regions. By controlling a Bayesian risk, our approach balances false positive and false negative risk control. Furthermore, it is based on a generative model that accounts for the spatial uncertainty on the localization of individual effects, due to spatial normalization errors. On both simulated and real fMRI datasets, we show that this new paradigm corrects several biases of the SPM-like approach, which either swells or misses the different active regions, depending on the choice of a cluster-forming threshold.
△ Less
Submitted 18 May, 2010;
originally announced May 2010.
-
Hypergeometric solutions for third order linear ODEs
Authors:
Edgardo S. Cheb-Terrab,
Austin D. Roche
Abstract:
In this paper we present a decision procedure for computing pFq hypergeometric solutions for third order linear ODEs, that is, solutions for the classes of hypergeometric equations constructed from the 3F2, 2F2, 1F2 and 0F2 standard equations using transformations of the form {x -> F(x), y -> P(x) y}, where F(x) is rational in x and P(x) is arbitrary. A computer algebra implementation of this wo…
▽ More
In this paper we present a decision procedure for computing pFq hypergeometric solutions for third order linear ODEs, that is, solutions for the classes of hypergeometric equations constructed from the 3F2, 2F2, 1F2 and 0F2 standard equations using transformations of the form {x -> F(x), y -> P(x) y}, where F(x) is rational in x and P(x) is arbitrary. A computer algebra implementation of this work is present in Maple 12.
△ Less
Submitted 14 April, 2008; v1 submitted 24 March, 2008;
originally announced March 2008.
-
An Abel ordinary differential equation class generalizing known integrable classes
Authors:
E. S. Cheb-Terrab,
A. D. Roche
Abstract:
We present a multi-parameter non-constant-invariant class of Abel ordinary differential equations with the following remarkable features. This one class is shown to unify, that is, contain as particular cases, all the integrable classes presented by Abel, Liouville and Appell, as well as all those shown in Kamke's book and various other references. In addition, the class being presented includes…
▽ More
We present a multi-parameter non-constant-invariant class of Abel ordinary differential equations with the following remarkable features. This one class is shown to unify, that is, contain as particular cases, all the integrable classes presented by Abel, Liouville and Appell, as well as all those shown in Kamke's book and various other references. In addition, the class being presented includes other new and fully integrable subclasses, as well as the most general parameterized class of which we know whose members can systematically be mapped into Riccati equations. Finally, many integrable members of this class can be systematically mapped into an integrable member of a different class. We thus find new integrable classes from previously known ones.
△ Less
Submitted 23 February, 2004; v1 submitted 8 February, 2000;
originally announced February 2000.
-
Integrating factors for second order ODEs
Authors:
E. S. Cheb-Terrab,
A. D. Roche
Abstract:
A systematic algorithm for building integrating factors of the form mu(x,y), mu(x,y') or mu(y,y') for second order ODEs is presented. The algorithm can determine the existence and explicit form of the integrating factors themselves without solving any differential equations, except for a linear ODE in one subcase of the mu(x,y) problem. Examples of ODEs not having point symmetries are shown to b…
▽ More
A systematic algorithm for building integrating factors of the form mu(x,y), mu(x,y') or mu(y,y') for second order ODEs is presented. The algorithm can determine the existence and explicit form of the integrating factors themselves without solving any differential equations, except for a linear ODE in one subcase of the mu(x,y) problem. Examples of ODEs not having point symmetries are shown to be solvable using this algorithm. The scheme was implemented in Maple, in the framework of the "ODEtools" package and its ODE-solver. A comparison between this implementation and other computer algebra ODE-solvers in tackling non-linear examples from Kamke's book is shown.
△ Less
Submitted 8 February, 2000;
originally announced February 2000.
-
Abel ODEs: Equivalence and Integrable Classes
Authors:
E. S. Cheb-Terrab,
A. D. Roche
Abstract:
A classification, according to invariant theory, of non-constant invariant Abel ODEs known as solvable and found in the literature is presented. A set of new integrable classes depending on one or no parameters, derived from the analysis of the works by Abel, Liouville and Appell, is also shown. Computer algebra routines were developed to solve ODEs members of these classes by solving their rela…
▽ More
A classification, according to invariant theory, of non-constant invariant Abel ODEs known as solvable and found in the literature is presented. A set of new integrable classes depending on one or no parameters, derived from the analysis of the works by Abel, Liouville and Appell, is also shown. Computer algebra routines were developed to solve ODEs members of these classes by solving their related equivalence problem. The resulting library permits a systematic solving of Abel type ODEs in the Maple symbolic computing environment.
△ Less
Submitted 25 January, 2000;
originally announced January 2000.
-
Integrating Factors and ODE Patterns
Authors:
E. S. Cheb-Terrab,
A. D. Roche
Abstract:
A systematic algorithm for building integrating factors of the form mu(x,y') or mu(y,y') for non-linear second order ODEs is presented. When such an integrating factor exists, the algorithm determines it without solving any differential equations. Examples of ODEs not having point symmetries are shown to be solvable using this algorithm. The scheme was implemented in Maple, in the framework of t…
▽ More
A systematic algorithm for building integrating factors of the form mu(x,y') or mu(y,y') for non-linear second order ODEs is presented. When such an integrating factor exists, the algorithm determines it without solving any differential equations. Examples of ODEs not having point symmetries are shown to be solvable using this algorithm. The scheme was implemented in Maple, in the framework of the ODEtools package and its ODE-solver. A comparison between this implementation and other computer algebra ODE-solvers in tackling non-linear examples from Kamke's book is shown.
△ Less
Submitted 1 June, 1998; v1 submitted 26 November, 1997;
originally announced November 1997.
