Neural network biased corrections: Cautionary study in background corrections for quenched jets

Summary

Neural networks (NNs) can be trained to correct for background fluctuations in jet measurements, but they introduce biases when applied to quenched jets. These biases are significant and dependent on jet-pT, affecting the accuracy of jet quenching measurements.

Highlights

• NNs can correct for background fluctuations in jet measurements.
• Biases are introduced when NNs are applied to quenched jets.
• Biases are significant and dependent on jet-pT.
• Jet quenching measurements are affected by these biases.
• NNs trained on pp jets embedded in hydro backgrounds exhibit biases.
• Quenching in bricks of QGP is used to mimic quenching in hydro events.
• RLeadJet values are calculated to demonstrate the effect of biases.

Key Insights

• NN background corrections introduce biases: The use of NNs to correct for background fluctuations in jet measurements introduces biases when applied to quenched jets. These biases are significant and dependent on jet-pT.
• Biases affect jet quenching measurements: The biases introduced by NN background corrections affect the accuracy of jet quenching measurements, which are crucial for understanding the properties of the quark-gluon plasma.
• NNs trained on pp jets exhibit biases: NNs trained on pp jets embedded in hydro backgrounds exhibit biases when applied to quenched jets, highlighting the need for more sophisticated training methods.
• Quenching in bricks of QGP mimics hydro events: Quenching in bricks of QGP is used to mimic quenching in hydro events, allowing for a more realistic simulation of jet quenching.
• RLeadJet values demonstrate bias effect: RLeadJet values are calculated to demonstrate the effect of biases on jet quenching measurements, showing that the biases can lead to significant errors.
• NNs can be improved with more sophisticated training: The biases introduced by NN background corrections can be mitigated with more sophisticated training methods, such as using quenched jets in the training dataset.
• Understanding biases is crucial for accurate measurements: Understanding the biases introduced by NN background corrections is crucial for accurate measurements of jet quenching and the properties of the quark-gluon plasma.

Mindmap

Citation

Stewart, D., & Putschke, J. (2024). Neural network biased corrections: Cautionary study in background corrections for quenched jets (Version 1). arXiv. https://doi.org/10.48550/ARXIV.2412.15440