De Finetti

De Finetti's theorem, focusing on exchangeable data (data whose probability distribution is invariant under permutations), provides a framework for understanding Bayesian inference without explicitly specifying prior distributions. Current research explores applications of this framework in diverse areas, including improving image registration through efficient feature-based initialization, optimizing path planning for infrastructure inspection using algorithms like GATSBI, and recovering topic distributions from large language models by leveraging the exchangeability of topic structure. This work has significant implications for advancing uncertainty quantification in machine learning, improving the efficiency of complex algorithms, and enhancing the interpretability of large-scale models.

Papers

August 6, 2024

Exchangeable Sequence Models Can Naturally Quantify Uncertainty Over Latent Concepts
Naimeng Ye, Hanming Yang, Andrew Siah, Hongseok Namkoong
Context Learning Bayesian Inference Sequence to Sequence Posterior Inference Sequence Modeling Sequence Model De Finetti

July 4, 2024

POLAFFINI: Efficient feature-based polyaffine initialization for improved non-linear image registration
Antoine Legouhy, Ross Callaghan, Hojjat Azadbakht, Hui Zhang
Affine Registration De Finetti

June 24, 2024

GATSBI: An Online GTSP-Based Algorithm for Targeted Surface Bridge Inspection and Defect Detection
Harnaik Dhami, Charith Reddy, Vishnu Dutt Sharma, Troi Williams, Pratap Tokekar
Defect Detection 3D Occupancy Traveling Salesman Problem Voxel Wise Baseline Algorithm Surface Inspection Bridge Inspection De Finetti

December 21, 2023

Deep de Finetti: Recovering Topic Distributions from Large Language Models
Liyi Zhang, R. Thomas McCoy, Theodore R. Sumers, Jian-Qiao Zhu, Thomas L. Griffiths
Large Language Model Latent Dirichlet Allocation Topic Distribution Probabilistic Topic De Finetti Topic Structure

October 5, 2022

Extending Conformal Prediction to Hidden Markov Models with Exact Validity via de Finetti's Theorem for Markov Chains
Buddhika Nettasinghe, Samrat Chatterjee, Ramakrishna Tipireddy, Mahantesh Halappanavar
Conformal Prediction Markov Chain Hidden Markov Model Markovian Sampling De Finetti

March 29, 2022

Causal de Finetti: On the Identification of Invariant Causal Structure in Exchangeable Data
Siyuan Guo, Viktor Tóth, Bernhard Schölkopf, Ferenc Huszár
Causal Discovery Person Identification Causal Structure Conditional Independence Causal Discovery Method Causal Discovery Algorithm Invariant Causal Feature De Finetti

March 12, 2022

GATSBI: Generative Adversarial Training for Simulation-Based Inference
Poornima Ramesh, Jan-Matthis Lueckmann, Jan Boelts, Álvaro Tejero-Cantero, David S. Greenberg, Pedro J. Gonçalves, Jakob H. Macke
Generative Adversarial Network Generative Adversarial Simulation Based Inference Implicit Prior De Finetti

De Finetti

Papers

Exchangeable Sequence Models Can Naturally Quantify Uncertainty Over Latent Concepts

POLAFFINI: Efficient feature-based polyaffine initialization for improved non-linear image registration

GATSBI: An Online GTSP-Based Algorithm for Targeted Surface Bridge Inspection and Defect Detection

Deep de Finetti: Recovering Topic Distributions from Large Language Models

Extending Conformal Prediction to Hidden Markov Models with Exact Validity via de Finetti's Theorem for Markov Chains

Causal de Finetti: On the Identification of Invariant Causal Structure in Exchangeable Data

GATSBI: Generative Adversarial Training for Simulation-Based Inference