-
Diffusion posterior sampling for simulation-based inference in tall data settings
Authors:
Julia Linhart,
Gabriel Victorino Cardoso,
Alexandre Gramfort,
Sylvain Le Corff,
Pedro L. C. Rodrigues
Abstract:
Determining which parameters of a non-linear model best describe a set of experimental data is a fundamental problem in science and it has gained much traction lately with the rise of complex large-scale simulators. The likelihood of such models is typically intractable, which is why classical MCMC methods can not be used. Simulation-based inference (SBI) stands out in this context by only requiri…
▽ More
Determining which parameters of a non-linear model best describe a set of experimental data is a fundamental problem in science and it has gained much traction lately with the rise of complex large-scale simulators. The likelihood of such models is typically intractable, which is why classical MCMC methods can not be used. Simulation-based inference (SBI) stands out in this context by only requiring a dataset of simulations to train deep generative models capable of approximating the posterior distribution that relates input parameters to a given observation. In this work, we consider a tall data extension in which multiple observations are available to better infer the parameters of the model. The proposed method is built upon recent developments from the flourishing score-based diffusion literature and allows to estimate the tall data posterior distribution, while simply using information from a score network trained for a single context observation. We compare our method to recently proposed competing approaches on various numerical experiments and demonstrate its superiority in terms of numerical stability and computational cost.
△ Less
Submitted 7 June, 2024; v1 submitted 11 April, 2024;
originally announced April 2024.
-
L-C2ST: Local Diagnostics for Posterior Approximations in Simulation-Based Inference
Authors:
Julia Linhart,
Alexandre Gramfort,
Pedro L. C. Rodrigues
Abstract:
Many recent works in simulation-based inference (SBI) rely on deep generative models to approximate complex, high-dimensional posterior distributions. However, evaluating whether or not these approximations can be trusted remains a challenge. Most approaches evaluate the posterior estimator only in expectation over the observation space. This limits their interpretability and is not sufficient to…
▽ More
Many recent works in simulation-based inference (SBI) rely on deep generative models to approximate complex, high-dimensional posterior distributions. However, evaluating whether or not these approximations can be trusted remains a challenge. Most approaches evaluate the posterior estimator only in expectation over the observation space. This limits their interpretability and is not sufficient to identify for which observations the approximation can be trusted or should be improved. Building upon the well-known classifier two-sample test (C2ST), we introduce L-C2ST, a new method that allows for a local evaluation of the posterior estimator at any given observation. It offers theoretically grounded and easy to interpret -- e.g. graphical -- diagnostics, and unlike C2ST, does not require access to samples from the true posterior. In the case of normalizing flow-based posterior estimators, L-C2ST can be specialized to offer better statistical power, while being computationally more efficient. On standard SBI benchmarks, L-C2ST provides comparable results to C2ST and outperforms alternative local approaches such as coverage tests based on highest predictive density (HPD). We further highlight the importance of local evaluation and the benefit of interpretability of L-C2ST on a challenging application from computational neuroscience.
△ Less
Submitted 9 October, 2023; v1 submitted 6 June, 2023;
originally announced June 2023.
-
Validation Diagnostics for SBI algorithms based on Normalizing Flows
Authors:
Julia Linhart,
Alexandre Gramfort,
Pedro L. C. Rodrigues
Abstract:
Building on the recent trend of new deep generative models known as Normalizing Flows (NF), simulation-based inference (SBI) algorithms can now efficiently accommodate arbitrary complex and high-dimensional data distributions. The development of appropriate validation methods however has fallen behind. Indeed, most of the existing metrics either require access to the true posterior distribution, o…
▽ More
Building on the recent trend of new deep generative models known as Normalizing Flows (NF), simulation-based inference (SBI) algorithms can now efficiently accommodate arbitrary complex and high-dimensional data distributions. The development of appropriate validation methods however has fallen behind. Indeed, most of the existing metrics either require access to the true posterior distribution, or fail to provide theoretical guarantees on the consistency of the inferred approximation beyond the one-dimensional setting. This work proposes easy to interpret validation diagnostics for multi-dimensional conditional (posterior) density estimators based on NF. It also offers theoretical guarantees based on results of local consistency. The proposed workflow can be used to check, analyse and guarantee consistent behavior of the estimator. The method is illustrated with a challenging example that involves tightly coupled parameters in the context of computational neuroscience. This work should help the design of better specified models or drive the development of novel SBI-algorithms, hence allowing to build up trust on their ability to address important questions in experimental science.
△ Less
Submitted 24 November, 2022; v1 submitted 17 November, 2022;
originally announced November 2022.
-
Numerical investigations of singularity formation in non-linear wave equations in the adiabatic limit
Authors:
Jean Marie Linhart
Abstract:
This dissertation deals with singularity formation in spherically symmetric solutions of the hyperbolic Yang Mills equations in (4+1) dimensions and in spherically symmetric solutions of C P^1 wave maps in (2+1) dimensions. These equations have known moduli spaces of time-independent (static) solutions. Evolution occurs close to the moduli space of static solutions. The evolution is modeled nume…
▽ More
This dissertation deals with singularity formation in spherically symmetric solutions of the hyperbolic Yang Mills equations in (4+1) dimensions and in spherically symmetric solutions of C P^1 wave maps in (2+1) dimensions. These equations have known moduli spaces of time-independent (static) solutions. Evolution occurs close to the moduli space of static solutions. The evolution is modeled numerically using an iterative finite differencing scheme, and modeling is done close to the adiabatic limit, i.e., with small velocities. The stability of the numerical scheme is analyzed and growth is shown to be bounded, yielding a convergence estimate for the numerical scheme. The trajectory of the approach is characterized, as well as the shape of the profile at any given time during the evolution.
△ Less
Submitted 30 May, 2001;
originally announced May 2001.
-
Fast and Slow Blowup in the S^2 Sigma Model and (4+1)-Dimensional Yang-Mills Model
Authors:
Jean Marie Linhart,
Lorenzo A. Sadun
Abstract:
We study singularity formation in spherically symmetric solutions of the charge-one and charge-two sector of the (2+1)-dimensional S^2 sigma-model and the (4+1)-dimensional Yang-Mills model, near the adiabatic limit. These equations are non-integrable, and so studies are performed numerically on rotationally symmetric solutions using an iterative finite differencing scheme that is numerically st…
▽ More
We study singularity formation in spherically symmetric solutions of the charge-one and charge-two sector of the (2+1)-dimensional S^2 sigma-model and the (4+1)-dimensional Yang-Mills model, near the adiabatic limit. These equations are non-integrable, and so studies are performed numerically on rotationally symmetric solutions using an iterative finite differencing scheme that is numerically stable. We evaluate the accuracy of predictions made with the geodesic approximation. We find that the geodesic approximation is extremely accurate for the charge-two sigma-model and the Yang-Mills model, both of which exhibit fast blowup. The charge-one sigma-model exhibits slow blowup. There the geodesic approximation must be modified by applying an infrared cutoff that depends on initial conditions.
△ Less
Submitted 17 May, 2001;
originally announced May 2001.
-
Fast Blow Up in the (4+1)-dimensional Yang Mills Model and the(2+1)-dimensional S^2 Sigma Model
Authors:
Jean Marie Linhart
Abstract:
We study singularity formation in spherically symmetric solitons of the (4+1) dimensional Yang Mills model and the charge two sector of the (2+1) dimensional S^2 sigma model, also known as $\IC P^1$ wave maps, in the adiabatic limit. These two models are very similar. Studies are performed numerically on radially symmetric solutions using an iterative finite differencing scheme. Predictions for…
▽ More
We study singularity formation in spherically symmetric solitons of the (4+1) dimensional Yang Mills model and the charge two sector of the (2+1) dimensional S^2 sigma model, also known as $\IC P^1$ wave maps, in the adiabatic limit. These two models are very similar. Studies are performed numerically on radially symmetric solutions using an iterative finite differencing scheme. Predictions for the evolution toward a singularity are made from an effective Lagrangian and confirmed numerically. In both models a characterization of the shape of a time slice f(r,T) with T fixed is provided, and ultimately yields an new approximate solutions to the differential equations that becomes exact in the adiabatic limit.
△ Less
Submitted 12 September, 1999;
originally announced September 1999.
-
Slow Blow Up in the (2+1)-dimensional S^2 Sigma Model
Authors:
Jean Marie Linhart
Abstract:
We study singularity formation in spherically symmetric solitons of the charge one sector of the (2+1) dimensional S^2 sigma model, also known as $\IC P^1$ wave maps, in the adiabatic limit. These equations are non-integrable, and so studies are performed numerically on radially symmetric solutions using an iterative finite differencing scheme. Analytic estimates are made by using an effective L…
▽ More
We study singularity formation in spherically symmetric solitons of the charge one sector of the (2+1) dimensional S^2 sigma model, also known as $\IC P^1$ wave maps, in the adiabatic limit. These equations are non-integrable, and so studies are performed numerically on radially symmetric solutions using an iterative finite differencing scheme. Analytic estimates are made by using an effective Lagrangian cutoff outside a ball of fixed radius. We show the geodesic approximation is valid when the cutoff is applied, with the cutoff approaching infinity linearly as the reciprocal of the initial velocity. Additionally a characterization of the shape of a time slice f(r,T) with T fixed is provided.
△ Less
Submitted 12 September, 1999;
originally announced September 1999.
