Abstract:We propose a deep learning method to build an AdS/QCD model from the data of hadron spectra. A major problem of generic AdS/QCD models is that a large ambiguity is allowed for the bulk gravity metric with which QCD observables are holographically calculated. We adopt the experimentally measured spectra of $\rho$ and $a_2$ mesons as training data, and perform a supervised machine learning which determines concretely a bulk metric and a dilaton profile of an AdS/QCD model. Our deep learning (DL) architecture is based on the AdS/DL correspondence (<a class="link-https" data-arxiv-id="1802.08313" href="https://arxiv.org/abs/1802.08313">arXiv:1802.08313</a>) where the deep neural network is identified with the emergent bulk spacetime.
High Energy Physics - Theory,Disordered Systems and Neural Networks,High Energy Physics - Phenomenology
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to construct an AdS/QCD model through the meson spectrum data measured by experiments, especially to determine the bulk metric and dilaton profile in this model. There are large uncertainties in the traditional AdS/QCD model, especially when choosing an appropriate bulk metric. The author adopts the method of supervised learning, uses the experimental data of ρ - meson and a2 - meson as the training set, and optimizes the model parameters through deep - learning techniques, so as to obtain an AdS/QCD model that can accurately describe these meson spectra.
Specifically, a deep - learning - based method is proposed in the paper, which is achieved through the following steps:
1. **Construct a neural network**: Transform the field equations in the AdS/QCD model into a deep neural network, where the weights of the network correspond to the metrics and dilaton fields in the model.
2. **Prepare the data set**: Use the experimentally measured mass spectra of ρ - meson and a2 - meson as input data, and generate positive samples and negative samples according to these data. Positive samples correspond to the experimentally measured meson masses, and negative samples are randomly generated mass values that do not match the experimental data.
3. **Train the neural network**: Train the neural network through the supervised learning method, so that the network can learn the correct metric and dilaton field configurations from the input meson mass data. A physically regularized term is introduced during the training process to ensure that the learned metric and dilaton field have physical significance.
4. **Verify and apply**: First, verify through the known soft - wall model data to confirm the effectiveness of the method. Then, use the actual experimental data for training to obtain the final AdS/QCD model parameters and discuss their physical significance.
Through this method, the author hopes to systematically determine the gravitational dual model from a given set of QCD correlation function data, thereby providing a new data - driven method for the study of strongly - coupled QCD.