Guillaume Staerman

Publications and Preprints

Point Processes and Physiological signals (MEG, EEG, ECG)

• Unmixing Noise from Hawkes Process to Model Learned Physiological Events (Submitted)
  Guillaume Staerman, Virginie Loison, Thomas Moreau  [arxiv]

• Flexible Parametric Inference for Space-Time Hawkes Processes (Submitted)
  Emilia Siviero, Guillaume Staerman, Stéphan Clémençon, Thomas Moreau  [arxiv]

• FaDIn: Fast Discretized Inference for Hawkes Processes with General Parametric Kernels (ICML 2023)
  Guillaume Staerman, Cédric Allain, Alexandre Gramfort, Thomas Moreau  [arxiv]

Robust Inference

• Hypothesis Transfer Learning with Surrogate Classification Losses: Generalization Bounds through Algorithmic Stability (ICML 2023)
  Anass Aghbalou, Guillaume Staerman  [arxiv]

• A Pseudo-Metric between Probability Distributions based on Depth-Trimmed Regions (TMLR, 2024)
  Guillaume Staerman, Pavlo Mozharovskyi, Stephan Clémençon, Florence d'Alché-Buc  [arxiv, code]

• Affine-Invariant Integrated Rank-Weighted Depth: Definition, Properties and Finite Sample Analysis (Electronic Journal of Statistics, 2023)
  Guillaume Staerman, Pavlo Mozharovskyi, Stephan Clémençon  [paper, code]

• Generalization Bounds in the Presence of Outliers: a Median-of-Means Study (ICML 2021)
  Pierre Laforgue, Guillaume Staerman, Stephan Clémençon  [arxiv]

• When OT meets MoM: Robust estimation of Wasserstein Distance (AISTATS 2021)
  Guillaume Staerman, Pierre Laforgue, Pavlo Mozharovskyi, Florence d'Alché-Buc  [arxiv, code]

• The Area of the Convex Hull of Sampled Curves: a Robust Functional Statistical Depth Measure (AISTATS 2020, oral)
  Guillaume Staerman, Pavlo Mozharovskyi, Stephan Clémençon  [arxiv, code]

Anomaly Detection

• Signature Isolation Forest (AISTATS 2025)
  Guillaume Staerman, Marta Campi, Gareth W. Peters, Tomoko Matsui  [arxiv]

• Unsupervised Layer-wise Score Aggregation for Textual OOD Detection (AAAI 2024)
  Maxime Darrin, Guillaume Staerman, Eduardo Gomes, Jackie CK Cheung, Pablo Piantanida, Pierre Colombo  [arxiv]

• A Functional Perspective on Multi-Layer Out-of-Distribution Detection (Submitted)
  Eduardo Gomes, Pierre Colombo, Guillaume Staerman, Nathan Noiry, Pablo Piantanida  [openreview]

• Toward Stronger Textual Attack Detectors (EMNLP 2023)
  Pierre Colombo, Guillaume Staerman, Marine Picot, Nathan Noiry, Pablo Piantanida  [paper]

• A Halfspace-Mass Depth-Based Method for Adversarial Attack Detection (TMLR, 2023)
  Marine Picot, Frederica Granese, Guillaume Staerman, Marco Romanelli, Francisco Messina, Pablo Piantanida, Pierre Colombo  [openreview]

• A Simple Unsupervised Data Depth-based Method to Detect Adversarial Images (Submitted)
  Marine Picot, Guillaume Staerman, Nathan Noiry, Francisco Messina, Pablo Piantanida, Pierre Colombo  [openreview]

• Beyond Mahalanobis-based Scores for Textual OOD Detection (NeurIPS 2022)
  Pierre Colombo, Eduardo Gomes, Guillaume Staerman, Nathan Noiry, Pablo Piantanida  

• Functional Anomaly Detection: a Benchmark Study (JDSA 2022)
  Guillaume Staerman, Eric Adjakossa, Pavlo Mozharovskyi, Vera Hofer, Jayant Sen Gupta, Stephan Clémençon  [arxiv]

• Functional Isolation Forest (ACML 2019)
  Guillaume Staerman, Pavlo Mozharovskyi, Stephan Clémençon, Florence d'Alché-Buc  [arxiv, code]

Applications to NLP

• Enhanced Hallucination Detection in Neural Machine Translation through Simple Detector Aggregation (EMNLP 2024)
  Anas Himmi, Guillaume Staerman, Marine Picot, Pierre Colombo, Nuno Miguel Guerreiro  [arxiv]

• A Novel Information Theoretic Objective to Disentangle Representations for Fair Textual Classification (AACL 2023)
  Pierre Colombo, Guillaume Staerman, Nathan Noiry, Pablo Piantanida  [arxiv]

• Learning Disentangled Textual Representations via Statistical Measures of Similarity (ACL 2022, oral)
  Pierre Colombo, Guillaume Staerman, Nathan Noiry, Pablo Piantanida  [arxiv]

• Automatic Text Evaluation through the Lens of Wasserstein Barycenters (EMNLP 2021, oral)
  Pierre Colombo, Guillaume Staerman, Chloé Clavel, Pablo Piantanida  [arxiv, code]