Simulation-based inference for neutrino interaction model parameter tuning

TL;DR

The paper tackles the challenge of tuning neutrino interaction models in GENIE given imperfect theory and growing data complexity. It applies simulation-based inference (SBI) with a neural posterior estimator (NPE) and a Masked Autoregressive Flow (MAF) to map predicted histograms back to the four MicroBooNE-tuned parameters $\theta_1$, $\theta_2$, $\theta_3$, and $\theta_4$. Using a training set of 200,000 GENIE-NUISANCE configurations and a 1,000-event test set, the SBI model recovers the true parameter values with unbiased posteriors and good calibration. The approach offers fast, amortized inference suitable for real data applications and holds promise for improving the efficiency of future neutrino-tuning campaigns, including potential application to the T2K data and comparisons with traditional likelihood fits.

Abstract

High-energy physics experiments studying neutrinos rely heavily on simulations of their interactions with atomic nuclei. Limitations in the theoretical understanding of these interactions typically necessitate ad hoc tuning of simulation model parameters to data. Traditional tuning methods for neutrino experiments have largely relied on simple algorithms for numerical optimization. While adequate for the modest goals of initial efforts, the complexity of future neutrino tuning campaigns is expected to increase substantially, and new approaches will be needed to make progress. In this paper, we examine the application of simulation-based inference (SBI) to the neutrino interaction model tuning for the first time. Using a previous tuning study performed by the MicroBooNE experiment as a test case, we find that our SBI algorithm can correctly infer the tuned parameter values when confronted with a mock data set generated according to the MicroBooNE procedure. This initial proof-of-principle illustrates a promising new technique for next-generation simulation tuning campaigns for the neutrino experimental community.

Figures (1)

• Figure 1: Upper left panel: Inferred posterior distributions of four parameters for a single event in the test data set. The gray dashed lines indicate the true values. Upper right panel: Residuals of four parameters for 1000 test events. The gray dashed lines indicate the true value. Lower left panel: Posterior coverage of the $\theta_i$ parameters for $1000$ test events. The diagonal black-dashed line indicates perfect uncertainty calibration. The gray regions indicate thresholds of 10% (dark gray) and 20% (light gray) uncertainty miscalibration. Lower right panel: The red points represent the MicroBooNE fit parameters reported in Ref. microboonegenietune and used to generate the test histogram $x_i$. In blue, we show the four parameters $\theta_i$ along with their corresponding $1\sigma$ error bars inferred by our network with $x_i$ as input. In orange, we show the prior ranges used to train the SBI. We observe an excellent match between the inferred and true parameter values.