(英) |
Recently, Alves et al. succeeded a discrimination among Haar measure, 1-design, and 2-design by learning out of time ordered correlators (OTOC) with constitutional neural networks (CNN).
In this study, we investigate whether discriminative models constructed by Alves et al.'s method and those constructed by other methods which are derived by modifying Alves et al's method can classify learning data solely depending on degrees of unitary designs.
For this purpose, we construct discriminative models by learning OTOC of the Haar measure and Random Clifford circuits with those different types of methods, and then, verify whether datasets generated by local random circuits with various different depth are classified
solely depending on their degree of designs.
As a result, we show that Alves et al.'s method does not discriminate data depending on its degree of design, but a method derived by modifying Alves et al's does. |