We introduce a model of hadronization based on invertible neural networks that faithfully reproduces a simplified version of the Lund string model for meson hadronization. Additionally, we introduce a new training method for normalizing flows, termed MAGIC, that improves the agreement between simulated and experimental distributions of high-level (macroscopic) observables by adjusting single-emission (microscopic) dynamics. Our results constitute an important step toward realizing a machine-learning based model of hadronization that utilizes experimental data during training. Finally, we demonstrate how a Bayesian extension to this normalizing-flow architecture can be used to provide analysis of statistical and modeling uncertainties on the generated observable distributions.
Author indications on fulfilling journal expectations
Provide a novel and synergetic link between different research areas.
Open a new pathway in an existing or a new research direction, with clear potential for multi-pronged follow-up work
Detail a groundbreaking theoretical/experimental/computational discovery
Present a breakthrough on a previously-identified and long-standing research stumbling block
Author comments upon resubmission
We thank the reviewers for the detailed reading of our manuscript. We have updated our manuscript with a number of changes based on their suggestions.
List of changes
- Included additional details and clarifying remarks on comparisons to cited work within the introduction, as suggested by the referees. - Added a footnote on page 3 providing a qualitative remark on the speed of the presented hadronization model compared to modern event generators. - Added a quantitative remark on the speed of re-weighting versus re-simulating during fine-tuning in section 4. - Added an expanded discussion regarding multi-dimensional MAGIC fine-tuning in the second to last paragraph of section 4. - Included additional references and an extended discussion in section 5 on Bayesian error estimation and its ability to quantify uncertainties related to the under- and over-fitting of training data. - Provided additional details on the binning used in the right panel of figure 4 within the figure caption, as suggested by the referees.
