Leveraging Reinforcement Learning, Genetic Algorithms and Transformers for background determination in particle physics

TL;DR

A novel approach that utilises Reinforcement Learning (RL) to overcome the aforementioned challenges by systematically determining the critical backgrounds affecting beauty hadron decay measurements is presented.

Abstract

Experimental studies of beauty hadron decays face significant challenges due to a wide range of backgrounds arising from the numerous possible decay channels with similar final states. For a particular signal decay, the process for ascertaining the most relevant background processes necessitates a detailed analysis of final state particles, potential misidentifications, and kinematic overlaps, which, due to computational limitations, is restricted to the simulation of only the most relevant backgrounds. Moreover, this process typically relies on the physicist's intuition and expertise, as no systematic method exists. This paper has two primary goals. First, from a particle physics perspective, we present a novel approach that utilises Reinforcement Learning (RL) to overcome the aforementioned challenges by systematically determining the critical backgrounds affecting beauty hadron decay measurements. While beauty hadron physics serves as the case study in this work, the proposed strategy is broadly adaptable to other types of particle physics measurements. Second, from a Machine Learning perspective, we introduce a novel algorithm which exploits the synergy between RL and Genetic Algorithms (GAs) for environments with highly sparse rewards and a large trajectory space. This strategy leverages GAs to efficiently explore the trajectory space and identify successful trajectories, which are used to guide the RL agent's training. Our method also incorporates a transformer architecture for the RL agent to handle token sequences representing decays.

Figures (11)

• Figure 1: Diagram showing the workflow of the algorithm.
• Figure 2: Example token sequence representing a state of the RL environment. Tokens have been highlighted in colors: particle tokens in yellow, tokens referencing the next element in the sequence in green, ")" and "End" tokens in red, and tokens not representing actions in blue. The signal is $B^0\to K^+ \pi^- \pi^0$ and the background generated by the agent $B^+\to K^{*+}( \ K^+ \pi^0 ) \ \pi^+ K^-$.
• Figure 4: Illustration of crossover (a) and mutation (b) processes for a tree-based gene representation. The relevant genes have been highlighted in colours. Mother particles are denoted by $M$, intermediate particles by $I$, and final state particles by $F$. Here, superscripts indicate the electric charge: $+$ for positively charged particles, $-$ for negatively charged particles and $0$ for neutral ones. Physically meaningful examples would be a crossover between $B^+\to K^{*0}(\to K^+ \pi^-) \pi^+ \gamma$ and $B^+\to K^+ \pi^+ \pi^0 \pi^-$, which produces an offspring of $B^+\to K^{*0}(\to K^+ \pi^-) \pi^+ \pi^0$ and $B^+\to K^+ \pi^+ \gamma \ \pi^-$, and a mutation transforming $B^+\to K^*(1680)^0(\to K^{*0}(\to K^+ \pi^-) \pi^0) \pi^+$ into $B^+\to K^*(1680)^0(\to K^{*0}(\to K^+ \pi^-) \pi^0) K^+$.
• Figure 5: Illustration of the random immigration variation process. A redundant individual in the offspring is replaced by a newly generated individual.
• Figure 6: Illustration of the custom variation process to construct an intermediate resonance. Mother particles are denoted by $M$, intermediate particles by $I$, and final state particles by $F$. Here, superscripts indicate the electric charge: $+$ for positively charged particles, $-$ for negatively charged particles and $0$ for neutral ones. In the example, genes $I_1^0(\to F_2^+ F_3^-)$ and $F_4^0$ are combined to construct the intermediate resonance $I_2^0(\to I_1^0(\to F_2^+ F_3^-) F_4^0 )$. A physically meaningful case would be the creation of the background candidate $B^+\to K^*(1680)^0 (\to K^{*0}(\to K^+ \pi^-) \pi^0) \pi^+$ from the GA individual $B^+\to K^{*0}(\to K^+ \pi^-) \pi^+ \pi^0$.