Skip to content
Valentin KILIAN edited this page Jul 8, 2025 · 86 revisions

This reading group at Oxford Statistics covers area related to Bayesian (nonparametric) methods, network modelling, statistical machine learning, power-laws in empirical data and the analysis of large neural networks. Email François if you are interested in participating, but not in the group.

2025

Date Time Room Paper/Topic Presenter Notes
12/01/2025 12:00 Meeting Room 1 Some ideas from Temporal Graph Neural Tangent Kernel with Graphon-Guaranteed Valentin All - Neurips 2024
19/01/2025 12:00 Length Optimization in Conformal Prediction Stefano All - Neurips 2024
12:30 An Information Theoretic Perspective on Conformal Prediction Kia All - Neurips 2024
26/01/2025 12:00 Boosted Conformal Prediction Intervals François All - Neurips 2024
04/02/2025 12:00 Meeting Room 1 A PAC-Bayesian Approach to Generalization Bounds for Graph Neural Networks Valentin
11/02/2025 12:00 1.20 Data thinning for convolution-closed distributions François
18/02/2025 13:00 1.20/MR1 Prediction-Powered E-Values Stefano
25/02/2025 12:00 1.20/MR1 Optimal Conformal Prediction for Small Areas Kia
04/03/2025 12:00 1.20/MR1 A unified recipe for deriving (time-uniform) PAC-Bayes bounds up to page 21 Valentin
18/03/2025 12:00 1.20/MR1 Some connections between PPI, doubly robust machine learning, the Augmented Inverse Propensity Weighted Estimator, Neyman orthogonality and control variates. The following slides may be useful François
01/04/2025 12:00 1.20/MR1 Outcome-Informed Weighting for Robust ATE Estimation Linying Yang Guest Speaker
29/04/2025 12:00 1.20/MR1 Reliable Active Learning via Influence Functions Stefano
28/05/2025 11:00 1.20/MR1 Influence-Guided Diffusion for Dataset Distillation Kia
03/06/2025 12:00 1.20/MR1 Bayesian Inference for Vertex-Series-Parallel Partial Orders Valentin
10/06/2025 12:00 1.20/MR1 Model Collapse Demystified: The Case of Regression François
17/06/2025 12:00 1.20/MR1 Conformal Prediction Beyond the Seen: A Missing Mass Perspective for Uncertainty Quantification in Generative Models Kia
08/07/2025 12:00 1.20/MR1 Rényi Neural Process Stefano

2024

Date Time Room Paper/Topic Presenter Notes
12/01/2024 11:00 Meeting Room 1 Abide by the law and follow the flow François
11:30 A graphon-signal analysis of GNNs Valentin
12:00 Thin and Deep Gaussian processes Stefano
12:30 Revisiting Logistic-softmax Likelihood in Bayesian Meta-Learning for Few-Shot Classification Kia
16/01/2024 12:00 1.20 The Quantization Model of Neural Scaling François
23/01/2024 12:00 Meeting Room 1 Sequence Modeling with Multiresolution Convolutional Memory Stefano
06/02/2024 12:00 Meeting Room 1 Bayesian inference of network structure from unreliable data Valentin
20/02/2024 12:00 Meeting Room 1 Posterior Re-calibration for Imbalanced Datasets Kia
12/03/2024 12:00 Meeting Room 1 Active Statistical Inference x Prediction-powered inference François
16/04/2024 12:00 Meeting Room 1 Graphon Neural Networks and the Transferability of Graph Neural Networks Valentin
07/05/2024 12:00 Meeting Room 1 Flow matching for generative modeling Stefano
14/05/2024 12:00 Meeting Room 1 Prediction-Oriented Bayesian Active Learning Kia
24/05/2024 16:00 1.20 Game-theoretic Statistics and safe anytime-valid inference (Ramdas, Grunwald, Vovk, Shafer, StatScience 2023) François
11/06/2024 09:45 LG.04 Graphon games: A statistical framework for network games and interventions Valentin
18/06/2024 14:15 Meeting Room 1 Bayesian Prediction-Powered Inference Stefano
25/06/2024 12:00 Meeting Room 1 Minimum-Risk Recalibration of Classifiers Kia
01/10/2024 12:00 1.20 Tweedie’s Formula and Selection Bias François
22/10/2024 12:00 Meeting Room 1 Robust and Conjugate Gaussian Process Regression (Altamirano et al., 2024) Stefano
05/11/2024 12:00 Meeting Room 1 On Statistical Bias In Active Learning: How and When To Fix It Kia
19/11/2024 12:00 Meeting Room 1 PAC-Bayesian Adversarially Robust Generalization Bounds for GNNs Valentin
03/12/2024 12:00 Meeting Room 1 Revisiting Optimism and Model Complexity in the Wake of Overparameterized Machine Learning I will mostly focus on the material covered in these notes François

List of potentially interesting papers to read

Neurips 2023

https://neurips.cc/virtual/2023/papers.html?filter=titles

Large neural networks/NTK

Dataset Shift, imbalanced/extreme classification

Networks

Online learning

Bayesian nonparametrics, power-laws

Knowledge distillation

Sequence modelling

Prediction-powered inference

Generative modeling

Neurips 2024

Conformal prediction

Boosted Conformal Prediction Intervals https://nips.cc/virtual/2024/poster/95004

Large language model validity via enhanced conformal prediction methods https://nips.cc/virtual/2024/poster/95729

Robust Conformal Prediction Using Privileged Information https://nips.cc/virtual/2024/poster/93870 https://arxiv.org/abs/2406.05405

Length Optimization in Conformal Prediction https://nips.cc/virtual/2024/poster/96056

An Information Theoretic Perspective on Conformal Prediction https://nips.cc/virtual/2024/poster/94151

Conformal Prediction for Class-wise Coverage via Augmented Label Rank Calibration https://nips.cc/virtual/2024/poster/94658

Similarity-Navigated Conformal Prediction for Graph Neural Networks https://nips.cc/virtual/2024/poster/94023

NTK - large neural nets

Bounds for the smallest eigenvalue of the NTK for arbitrary spherical data of arbitrary https://nips.cc/virtual/2024/poster/93764

Shaving Weights with Occam's Razor: Bayesian Sparsification for Neural Networks using the Marginal Likelihood https://nips.cc/virtual/2024/poster/95142

Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization https://nips.cc/virtual/2024/poster/94399

GNNs

Temporal Graph Neural Tangent Kernel with Graphon-Guaranteed https://nips.cc/virtual/2024/poster/96832

What Is Missing For Graph Homophily? Disentangling Graph Homophily For Graph Neural Networks https://nips.cc/virtual/2024/poster/95879

Probabilistic Weather Forecasting with Hierarchical Graph Neural Networks https://nips.cc/virtual/2024/poster/93149

On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks https://nips.cc/virtual/2024/poster/96753

Other

Do Finetti: On Causal Effects for Exchangeable Data https://nips.cc/virtual/2024/poster/96631

Bayesian Nonparametrics Meets Data-Driven Distributionally Robust Optimization https://nips.cc/virtual/2024/poster/96385

A Unified Recipe for Deriving (Time-Uniform) PAC-Bayes Bounds https://nips.cc/virtual/2024/poster/98308

Quasi-Bayes meets Vines https://nips.cc/virtual/2024/poster/94131

https://nips.cc/virtual/2024/poster/95321 Scaling Laws in Linear Regression: Compute, Parameters, and Data

Bias Detection via Signaling https://nips.cc/virtual/2024/poster/96686

LLM Processes: Numerical Predictive Distributions Conditioned on Natural Language https://nips.cc/virtual/2024/poster/95832

Approximately Equivariant Neural Processes https://nips.cc/virtual/2024/poster/94315

Robust Gaussian Processes via Relevance Pursuit https://nips.cc/virtual/2024/poster/96603

Decomposable Transformer Point Processes https://nips.cc/virtual/2024/poster/95355

Is Score Matching Suitable for Estimating Point Processes? https://nips.cc/virtual/2024/poster/95838

Physics-Informed Variational State-Space Gaussian Processes https://nips.cc/virtual/2024/poster/93352