Quark Mass Models and Reinforcement Learning

Abstract

In this paper, we apply reinforcement learning to the problem of constructing models in particle physics. As an example environment, we use the space of Froggatt-Nielsen type models for quark masses. Using a basic policy-based algorithm we show that neural networks can be successfully trained to construct Froggatt-Nielsen models which are consistent with the observed quark masses and mixing. The trained policy networks lead from random to phenomenologically acceptable models for over 90% of episodes and after an average episode length of about 20 steps. We also show that the networks are capable of finding models proposed in the literature when starting at nearby configurations.

Exploring the flavor structure of quarks and leptons with reinforcement learning

“…The hyperparameters in -greedy method are described in the previous section. About the step number N step = 32, the same value was used in the previous research that focuses on only the quark sector [12]. In ref.…”

“…In ref. [12], it was shown that terminal states can be reached after a sufficient amount of learning. Therefore, it is expected that N step = 32 is enough to achieve terminal states in the current situation where the quark sector and the lepton sector are searched separately.…”

“…2 In the framework of RL, an agent autonomously discovers desirable behavior to solve given problems, where a systematic search is impossible. So far, such a technique was utilized to find the parameter space of FN models with an emphasis on the quark sector [12], where only the experimental values of quark masses and mixing angles are efficiently reproduced. However, it is quite important to see whether one can reproduce the flavor structure of all the fermion masses and mixings.…”

Solving Conformal Field Theories with Artificial Intelligence

“…However, unlike [11,18], instead of using direct numerical methods, we take advantage of recent developments in the field of Artificial Intelligence and Machine Learning, where processes dealing with large parameter spaces have been proven to be very effective. Our preferred type of algorithms are Reinforcement-Learning (RL) algorithms [20]; RL implementations have also recently appeared in the context of String Theory [21][22][23][24][25]. By default, these do not require externally provided data for training-they learn on their own through exploration.…”

Applying machine learning to the Calabi-Yau orientifolds with string vacua

“…This technique allows people to learn lots of quantities of Calabi-Yau manifolds, from its toric building blocks like the polytope structure [21,22] and triangulations [23,24], to the calculation of Hodge numbers [25][26][27][28], numerical metrics [29][30][31][32] and line bundle cohomologies [33,34]. Besides, machine learning has also been applied to study and find certain structures on Calabi-Yau for model building [35][36][37][38][39][40][41][42].…”