Cross-Platform Autonomous Control of Minimal Kitaev Chains

This folder contains the raw data and code used to generate the plots for the paper Cross-Platform Autonomous Control of Minimal Kitaev Chains. To run the Jupyter notebook, install Anaconda and execute: conda env create -f environment.yml followed by: conda activate KitaevML Finally, jupyter lab to launch the notebook. Raw data are stored in netCDF (.nc) format, or as full QCoDeS databases (.db). Logs of autonomous tuning runs are saved as log files (.txt). The experimental datasets are exported by the data acquisition package QCoDeS and can be read as an xarray Dataset. The ipython notebook Data labelling notebook.ipynb contains the code for labelling the 2DEG and nanowire datasets. The ipython notebook Zenodo Paper Figures.ipynb contains all the code for generating the paper figures. The machine learning model and automated tuning algorithm can be publicly accessed here: https://gitlab.com/QMAI/papers/crossplatformkitaev.This repo contains a pip-installable module ``PMMSGD` that can be used to predict (Delta-t)/(Delta+t) using: pred = PMMSGD.get_Delta_minus_t(dbx, 'CNN_weighted_ratio_theory', ['G_LL*G_RR']) which uses the neural network that was exclusively trained on theory data For the automated tuning algorithm we used this predictor: pred = PMMSGD.get_Delta_minus_t(dbx, 'CNN_weighted_ratio', ['G_LL*G_RR']) which uses the neural network that was additionally trained on data from the 2DEG device.本文件夹包含论文《最小Kitaev链的跨平台自主控制》绘制配图所需的原始数据与代码。 若需运行该Jupyter笔记本（Jupyter Notebook），请先安装Anaconda，并按顺序执行以下命令： conda env create -f environment.yml 随后执行： conda activate KitaevML 最后执行： jupyter lab 以启动该笔记本。 原始数据以netCDF（.nc）格式存储，或以完整QCoDeS数据库（.db）格式存储。自主调谐运行的日志保存为日志文件（.txt）。实验数据集由数据采集包QCoDeS导出，可作为xarray Dataset读取。 IPython笔记本`Data labelling notebook.ipynb`包含了对二维电子气（2DEG，Two-Dimensional Electron Gas）与纳米线数据集进行标注的代码；IPython笔记本`Zenodo Paper Figures.ipynb`包含了生成该论文所有配图的全部代码。 本机器学习模型与自动调谐算法可通过以下链接公开获取：https://gitlab.com/QMAI/papers/crossplatformkitaev。本仓库包含一个可通过pip安装的模块`PMMSGD`，可通过如下代码预测(Δ-t)/(Δ+t)： pred = PMMSGD.get_Delta_minus_t(dbx, 'CNN_weighted_ratio_theory', ['G_LL*G_RR']) 该接口使用了仅在理论数据上训练得到的神经网络。 针对自动调谐算法，我们采用了如下预测接口： pred = PMMSGD.get_Delta_minus_t(dbx, 'CNN_weighted_ratio', ['G_LL*G_RR']) 该接口使用了额外在二维电子气器件数据上训练得到的神经网络。