Flow Annealed Importance Sampling Bootstrap meets Differentiable Particle Physics

Abstract

High-energy physics requires the generation of large numbers of simulated data samples from complex but analytically tractable distributions called matrix elements. Surrogate models, such as normalizing flows, are gaining popularity for this task due to their computational efficiency. We adopt an approach based on Flow Annealed importance sampling Bootstrap (FAB) that evaluates the differentiable target density during training and helps avoid the costly generation of training data in advance. We show that FAB reaches higher sampling efficiency with fewer target evaluations in high dimensions in comparison to other methods.

Accepted at "Machine Learning: Science and Technology"

References in corpus (33)

• Adam: A Method for Stochastic Optimization
• Array Programming with NumPy
• The automated computation of tree-level and next-to-leading order differential cross sections, and their matching to parton shower simulations
• MadGraph 5 : Going Beyond
• Variational Inference with Normalizing Flows
• Event Generation with Sherpa 2.2
• Event Generation with Normalizing Flows
• Effective Sample Size for Importance Sampling based on discrepancy measures
• Adaptive Multidimensional Integration: VEGAS Enhanced
• i-flow: High-dimensional Integration and Sampling with Normalizing Flows
• Exploring phase space with Neural Importance Sampling
• Measurement of the t t-bar Production Cross Section in pp Collisions at 7 TeV in Lepton + Jets Events Using b-quark Jet Identification
• Accelerating Monte Carlo event generation -- rejection sampling using neural network event-weight estimates
• Neural Network-Based Approach to Phase Space Integration
• Rethinking the Effective Sample Size
• Resonance Production and $\pi \pi$ S-wave in ${\pi }^{-}+p\to {\pi }^{-}{\pi }^{-}{\pi }^{+}+p_{recoil}$ at 190 GeV/c
• Phase Space Sampling and Inference from Weighted Events with Autoregressive Flows
• MadNIS -- Neural Multi-Channel Importance Sampling
• Inclusive and differential cross-sections for dilepton $t\bar{t}$ production measured in $\sqrt{s}=13$TeV $pp$ collisions with the ATLAS detector
• The MadNIS Reloaded
• PDFFlow: parton distribution functions on GPU
• Further tests of lepton flavour universality from the charged lepton energy distribution in $b\to c$ semileptonic decays: The case of ${\Lambda }_b\to {\Lambda }_c\ell {\bar{\nu }}_{\ell }$
• Exhaustive Neural Importance Sampling applied to Monte Carlo event generation
• The measurable angular distribution of ${\Lambda }_b^0\to {\Lambda }_c^{+}(\to {\Lambda }^0{\pi }^{+}){\tau }^{-}(\to {\pi }^{-}{\nu }_{\tau }){\bar{\nu }}_{\tau }$ decay
• Amplitude analysis of the ${\Lambda }_c^{+}\to p{K}^{-}{\pi }^{+}$ decay and ${\Lambda }_c^{+}$ baryon polarization measurement in semileptonic beauty hadron decays
• Measurement of the $t\bar{t}$ production cross-section in $pp$ collisions at $\sqrt{s}=5.02$ TeV with the ATLAS detector
• Measurement of ttbar production in the tau + jets topology using ppbar collisions at sqrt{s} = 1.96 TeV
• Differentiable MadNIS-Lite
• Flow Annealed Importance Sampling Bootstrap
• Differentiable Matrix Elements with MadJax
• BlackJAX: Composable Bayesian inference in JAX
• Extracting maximum information from polarised baryon decays via amplitude analysis: the ${\Lambda }_c^{+}\to p{K}^{-}{\pi }^{+}$ case
• ${\Lambda }_{\mathrm{c}}^{+}$ polarimetry using the dominant hadronic mode

Cited by in corpus (2)

• Sampling NNLO QCD phase space with normalizing flows
• Accelerating multijet-merged event generation with neural network matrix element surrogates