• Publications
  • Influence
Learning explanations that are hard to vary
TLDR
It is shown that averaging gradients across examples -- akin to a logical OR of patterns -- can favor memorization and `patchwork' solutions that sew together different strategies, instead of identifying invariances.
Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style
TLDR
Causal3DIdent, a dataset of high-dimensional, visually complex images with rich causal dependencies, which is used to study the effect of data augmentations performed in practice, and numerical simulations with dependent latent variables are consistent with theory.
Relative gradient optimization of the Jacobian term in unsupervised deep learning
TLDR
Based on relative gradients, the matrix structure of neural network parameters is exploited to compute updates efficiently even in high-dimensional spaces; the computational cost of the training is quadratic in the input size, in contrast with the cubic scaling of the naive approaches.
Independent mechanism analysis, a new concept?
TLDR
This work provides theoretical and empirical evidence that its approach circumvents a number of nonidentifiability issues arising in nonlinear blind source separation, by thinking of each source as independently influencing the mixing process.
The Incomplete Rosetta Stone problem: Identifiability results for Multi-view Nonlinear ICA
TLDR
It is proved that independent latent sources with arbitrary mixing can be recovered as long as multiple, sufficiently different noisy views are available, in contrast to known identifiability results for nonlinear ICA.
Modeling Shared Responses in Neuroimaging Studies through MultiView ICA
TLDR
This work proposes a novel MultiView Independent Component Analysis model for group studies, where data from each subject are modeled as a linear combination of shared independent sources plus noise, and develops an alternate quasi-Newton method for maximizing the likelihood.
Simpson's Paradox in COVID-19 Case Fatality Rates: A Mediation Analysis of Age-Related Causal Effects
TLDR
Causal inference, in particular mediation analysis, can be used to resolve apparent statistical paradoxes; help educate the public and decision-makers alike; avoid unsound comparisons; and answer a range of causal questions pertaining to the pandemic, subject to transparently stated assumptions.
On Maximum Entropy and Inference
TLDR
This work explores the approach in the case of spin models with interactions of arbitrary order, and discusses how relevant interactions can be inferred, and illustrates the method's ability to recover the correct model in a few prototype cases.
Privacy-Preserving Causal Inference via Inverse Probability Weighting
TLDR
A novel framework for privacy-preserving IPW (PP-IPW) methods is provided, a theoretical analysis of the effects of the proposed privatisation procedure on the estimated average treatment effect is included, and the empirical results are consistent with the theoretical findings.
Causal Inference Through the Structural Causal Marginal Problem
TLDR
This work introduces an approach to counterfactual inference based on merging information from multiple datasets and formalises this approach for categorical SCMs using the response function formulation and shows that it reduces the space of allowed marginal and joint SCMs.
...
...