Solving PDEs with Generative AI: An Algorithm for Joint Inversion and Posterior Sampling

ETKİNLİK BAŞLANGIÇ TARİHİ
30 Nisan 2025 15:30
ETKİNLİK BİTİŞ TARİHİ
30 Nisan 2025 17:30
ETKİNLİK TÜRÜ
Seminer
Solving PDEs with Generative AI: An Algorithm for Joint Inversion and Posterior Sampling
Solving PDEs with Generative AI: An Algorithm for Joint Inversion and Posterior Sampling

Speaker:

Assoc. Prof. Çağrı Diner

Department of Management Information Systems is pleased to host a seminar by Assoc. Prof. Çağrı Diner from the Geophysics Department at Boğaziçi University, Kandilli Observatory and Earthquake Research Institute.

Our speaker will present his talk titled, "Solving PDEs with Generative AI: An Algorithm for Joint Inversion and Posterior Sampling" on Thursday, April 30, 2026, at 1:00 PM in IB412, Faculty of Economics and Administrative Sciences.

Abstract:
Partial differential equations (PDEs) are crucial for predicting physical systems, but inverse problems in geophysics are highly challenging and notoriously ill-posed. Traditional solvers are computationally expensive and often fail when applied to highly sparse observations, yielding only a single deterministic solution. This talk explores a modern alternative: denoising diffusion models. Dr. Diner will present how this class of generative AI can learn joint probability distributions to concurrently solve forward and inverse PDE problems. By guiding the reverse stochastic differential equation (SDE) with sparse observational data and PDEs, this approach generates a rich ensemble of valid Earth models rather than a single guess.
Seminar Details

Seminar Title: Solving PDEs with Generative AI: An Algorithm for Joint Inversion and Posterior Sampling

Speaker: Assoc. Prof. Çağrı Diner (Kandilli Observatory and Earthquake Research Institute)

Time/Date: 13:00 / 30.04.2026

Location: IB412, Faculty of Economics and Administrative Sciences