arXiv subjects

Frédérique Letué

Publications and source records attributed to Frédérique Letué.

8 recordsLinked to original sources

arXiv2026

Marked point processes intensity estimation using sparse group Lasso method applied to locations of lucrative and cooperative banks in mainland France

In this paper, we model the locations of five major banks in mainland France, two lucrative and three cooperative institutions based on socio-economic considerations. Locations of banks are collected using web scrapping and constitute a bivariate spatial point process for which we estimate nonparametrically summary functions (intensity, Ripley and cross-Ripley's K functions). This shows that the pattern is highly inhomogenenous and exhibits a clustering effect especially at small scales, and thus a significant departure to the bivariate (inhomogeneous) Poisson point process is pointed out. We also collect socio-economic datasets (at the living area level) from INSEE and propose a parametric modelling of the intensity function using these covariates. We propose a group-penalized bivariate composite likelihood method to estimate the model parameters, and we establish its asymptotic properties. The application of the methodology to the banking dataset provides new insights into the specificity of the cooperative model within the sector, particularly in relation to the theories of institutional isomorphism.

stat.ME↗

arXiv2026

Listen to the Features: Voice Anonymization Driven by Content Embedding Matching over Signal Reconstruction

The paper presents a voice anonymization model focusing on preserving content rather than producing realistic speech. It relies on content embeddings extracted from a frozen pretrained wav2vec2 encoder. These embeddings are decoded into an anonymized signal using vector quantization and a HiFi-GAN vocoder, both trained on LibriTTS without any waveform reconstruction loss or speaker embedding mapping. The training objective enforces that embeddings of the anonymized signal match those of the original one. While training, an auxiliary speaker classification branch with a gradient reversal layer is used to discard speakerspecific information. Results show that this straightforward embedding-based approach achieves very low WER (2.53) with an anonymization performance (EER 13.39) ranking within first level for VPC. Notably, emotions are partially preserved (UAR 43.91), even without a supporting training objective, while the anonymized voice is audible without reconstruction loss.

eess.SP↗

arXiv2026

Examining the robustness of a model selection procedure in the binary latent block model through a language placement test data set

When entering French university, the students' foreign language level is assessed through a placement test. In this work, we model the placement test results using binary latent block models which allow to simultaneously form homogeneous groups of students and of items. However, a major difficulty in latent block models is to select correctly the number of groups of rows and the number of groups of columns. The first purpose of this paper is to tune the number of initializations needed to limit the initial values problem in the estimation algorithm in order to propose a model selection procedure in the placement test context. Computational studies based on simulated data sets and on two placement test data sets are investigated. The second purpose is to investigate the robustness of the proposed model selection procedure in terms of stability of the students groups when the number of students varies.

stat.ME↗

arXiv2024

Exploring VQ-VAE with Prosody Parameters for Speaker Anonymization

Human speech conveys prosody, linguistic content, and speaker identity. This article investigates a novel speaker anonymization approach using an end-to-end network based on a Vector-Quantized Variational Auto-Encoder (VQ-VAE) to deal with these speech components. This approach is designed to disentangle these components to specifically target and modify the speaker identity while preserving the linguistic and emotionalcontent. To do so, three separate branches compute embeddings for content, prosody, and speaker identity respectively. During synthesis, taking these embeddings, the decoder of the proposed architecture is conditioned on both speaker and prosody information, allowing for capturing more nuanced emotional states and precise adjustments to speaker identification. Findings indicate that this method outperforms most baseline techniques in preserving emotional information. However, it exhibits more limited performance on other voice privacy tasks, emphasizing the need for further improvements.

cs.CV↗

arXiv2023

Multimodal Group Emotion Recognition In-the-wild Using Privacy-Compliant Features

This paper explores privacy-compliant group-level emotion recognition ''in-the-wild'' within the EmotiW Challenge 2023. Group-level emotion recognition can be useful in many fields including social robotics, conversational agents, e-coaching and learning analytics. This research imposes itself using only global features avoiding individual ones, i.e. all features that can be used to identify or track people in videos (facial landmarks, body poses, audio diarization, etc.). The proposed multimodal model is composed of a video and an audio branches with a cross-attention between modalities. The video branch is based on a fine-tuned ViT architecture. The audio branch extracts Mel-spectrograms and feed them through CNN blocks into a transformer encoder. Our training paradigm includes a generated synthetic dataset to increase the sensitivity of our model on facial expression within the image in a data-driven way. The extensive experiments show the significance of our methodology. Our privacy-compliant proposal performs fairly on the EmotiW challenge, with 79.24% and 75.13% of accuracy respectively on validation and test set for the best models. Noticeably, our findings highlight that it is possible to reach this accuracy level with privacy-compliant features using only 5 frames uniformly distributed on the video.

cs.AI↗

arXiv2022

Adaptive lasso and Dantzig selector for spatial point processes intensity estimation

Lasso and Dantzig selector are standard procedures able to perform variable selection and estimation simultaneously. This paper is concerned with extending these procedures to spatial point process intensity estimation. We propose adaptive versions of these procedures, develop efficient computational methodologies and derive asymptotic results for a large class of spatial point processes under an original setting where the number of parameters, i.e. the number of spatial covariates considered, increases with the expected number of data points. Both procedures are compared theoretically, in a simulation study, and in a real data example.

stat.ME↗

arXiv2017

Spatial point processes intensity estimation with a diverging number of covariates

Feature selection procedures for spatial point processes parametric intensity estimation have been recently developed since more and more applications involve a large number of covariates. In this paper, we investigate the setting where the number of covariates diverges as the domain of observation increases. In particular, we consider estimating equations based on Campbell theorems derived from Poisson and logistic regression likelihoods regularized by a general penalty function. We prove that, under some conditions, the consistency, the sparsity, and the asymptotic normality are valid for such a setting. We support the theoretical results by numerical ones obtained from simulation experiments and an application to forestry datasets.

stat.ME↗

arXiv2017

Convex and non-convex regularization methods for spatial point processes intensity estimation

This paper deals with feature selection procedures for spatial point processes intensity estimation. We consider regularized versions of estimating equations based on Campbell theorem derived from two classical functions: Poisson likelihood and logistic regression likelihood. We provide general conditions on the spatial point processes and on penalty functions which ensure consistency, sparsity and asymptotic normality. We discuss the numerical implementation and assess finite sample properties in a simulation study. Finally, an application to tropical forestry datasets illustrates the use of the proposed methods.

stat.ME↗