Fast Inference Using Automatic Differentiation and Neural Transport in Astroparticle Physics

Dorian W. P. Amaral,Shixiao Liang,Juehang Qin,Christopher Tunnell

2024-05-24

Abstract:Multi-dimensional parameter spaces are commonly encountered in astroparticle physics theories that attempt to capture novel phenomena. However, they often possess complicated posterior geometries that are expensive to traverse using techniques traditional to this community. Effectively sampling these spaces is crucial to bridge the gap between experiment and theory. Several recent innovations, which are only beginning to make their way into this field, have made navigating such complex posteriors possible. These include GPU acceleration, automatic differentiation, and neural-network-guided reparameterization. We apply these advancements to astroparticle physics experimental results in the context of novel neutrino physics and benchmark their performances against traditional nested sampling techniques. Compared to nested sampling alone, we find that these techniques increase performance for both nested sampling and Hamiltonian Monte Carlo, accelerating inference by factors of $\sim 100$ and $\sim 60$, respectively. As nested sampling also evaluates the Bayesian evidence, these advancements can be exploited to improve model comparison performance while retaining compatibility with existing implementations that are widely used in the natural sciences.

Machine Learning,High Energy Physics - Phenomenology

What problem does this paper attempt to address?

This paper focuses on exploring the problem of complex high-dimensional parameter space in astroparticle physics. These theories often involve a large number of parameters, and efficient methods are needed to handle multidimensional spaces with complex posterior geometries. Traditional methods such as Nested Sampling (NS) are inefficient when facing such problems. The paper proposes a method that combines the latest technologies, including GPU acceleration, automatic differentiation, and Neural Transport (NeuTra) guided reparameterization, to improve the inference performance in new neutrino physics. Through comparative experiments, these methods are applied to synthetic problems and parameterization of neutrino non-standard interactions (NSI), with benchmark tests against traditional nested sampling techniques. The research found that these techniques can significantly improve the efficiency of nested sampling and Hamiltonian Monte Carlo (HMC), accelerating them by about 100 times and 60 times, respectively. In addition, since nested sampling can also compute Bayesian evidence, these advancements can be used to improve model comparison performance while maintaining compatibility with existing implementations widely used in natural science research. The paper also introduces background knowledge, including the basic principles of nested sampling and HMC, as well as their applications in particle physics. Additionally, the paper discusses the role of neutrinos in astroparticle experiments, particularly the parameterization of neutrino non-standard interactions, and demonstrates how NeuTra can enhance sampling efficiency. In conclusion, the paper aims to address the challenge of inference in complex parameter spaces in astroparticle physics by introducing modern computational tools such as GPU acceleration and neural networks, thereby accelerating the bridge-building between theory and experimental data.

Fast Inference Using Automatic Differentiation and Neural Transport in Astroparticle Physics

Deep Learning and genetic algorithms for cosmological Bayesian inference speed-up

Exploring parameter spaces with artificial intelligence and machine learning black-box optimization algorithms

Differentiable Cosmological Simulation with Adjoint Method

Differentiable Multi-Fidelity Fusion: Efficient Learning of Physics Simulations with Neural Architecture Search and Transfer Learning

Accelerated Inference for Partially Observed Markov Processes using Automatic Differentiation

Approximate Differentiable Likelihoods for Astroparticle Physics Experiments

Fast posterior probability sampling with normalizing flows and its applicability in Bayesian analysis in particle physics

Efficient Compression of Sparse Accelerator Data Using Implicit Neural Representations and Importance Sampling

Learning functional priors and posteriors from data and physics

An implementation of neural simulation-based inference for parameter estimation in ATLAS

Accelerated Sampling of Rare Events using a Neural Network Bias Potential

A comparison of Bayesian sampling algorithms for high-dimensional particle physics and cosmology applications

Learning new physics efficiently with nonparametric methods

Fast inference of deep neural networks in FPGAs for particle physics

Fast Point Cloud Generation with Diffusion Models in High Energy Physics

Improving Neutrino Energy Reconstruction with Machine Learning

Differentiable Cosmological Simulation with the Adjoint Method

NeuTra-lizing Bad Geometry in Hamiltonian Monte Carlo Using Neural Transport

Neural networks-based variationally enhanced sampling