Dear all,
Season 4 of the One Word ABC Seminar https://warwick.ac.uk/oneworldabc is starting this week on Thursday the 29th September, 1.30pm UK time.
The first speaker of the season will be Till Hoffman, who will talk about "Minimizing the Expected Posterior Entropy Yields Optimal Summary Statistics", with an abstract reported below.
The link to join the talk is https://bristol-ac-uk.zoom.us/j/93353511660?pwd=MmFTNUR5QkFBeStlenNVemtZbzJy...
Meeting ID: 933 5351 1660 Passcode: 406075
We are looking forward to seeing you there, Best regards, Massimiliano on the behalf of the One World ABC Organisers
When: 29th September, 1.30pm UK time Speaker: Till Hoffmann https://tillahoffmann.github.io/ (Harward T.H. Chan School of Publich Health) Title: Minimizing the Expected Posterior Entropy Yields Optimal Summary Statistics Abstract: Extracting low-dimensional summary statistics from large datasets is essential for efficient (likelihood-free) inference. We propose obtaining summary statistics by minimizing the expected posterior entropy (EPE) under the prior predictive distribution of the model. We show that minimizing the EPE is equivalent to learning a conditional density estimator for the posterior as well as other information-theoretic approaches. Further summary extraction methods (including minimizing the L² Bayes risk, maximizing the Fisher information, and model selection approaches) are special or limiting cases of EPE minimization. We demonstrate that the approach yields high fidelity summary statistics by applying it to both a synthetic benchmark as well as a population genetics problem. We not only offer concrete recommendations for practitioners but also provide a unifying perspective for obtaining informative summary statistics. Reference: T. Hoffmann and J.P. Onnela. Minimizing the Expected Posterior Entropy Yields Optimal Summary Statistics. Preprint at ArXiv:2206.02340, 2022
------ Dr. Massimiliano Tamborrino Assistant Professor Department of Statistics University of Warwick https://warwick.ac.uk/tamborrino