Training data for ARPESNet

Datasets used for training the ARPESNet autoencoder.

Datasets:

The training and test data are stored in the .zip files. Unzipping these will provide the files created by saving PyTorch tensors. These can be loaded with the python code:

import torch

data = torch.load("filename.pt")

 where filename can be any of the files in this repository.

train_data.zip & test_data.zip: training and testing data, respectively. Collection of 256x256 ARPES spectra (images) obtained by randomly slicing 3D of 28 (train) and 18 (test) high resolution angle scans, covering 19 material systems: Au(110), Au(111), Bi2Se3(111), CoO2, on Au(111), CrSBr(001), Graphene on Ir(111), Graphene on Ru(0001), single-layer MoS2 on Au(111), single-layer NbSe2 on bilayer graphene, NdTe3(010), Pd(100), Pd(111), Pt(111), Rb-doped Bi2Se3(111), Ru(0001), P-δ-layer on Si(001), single-layer TaS2 on Au(111), single-layer WS2 on Ag(111) and single-layer WS2on Au(111).

Each file corresponds to one material system, for which 500 images were generated, resulting in a tensor of shape 500x256x256 each.

test_imgs.pt: 6 ARPES spectra used for visual inspection and performance test of the ARPESNet autoencoder.

cluster_centers.pt: ARPES spectra extracted by slicing an angle scan obtained measuring a Bi2Se3 crystal. These are used to generate simulated nanoARPES maps for testing clustering performance.

dataset_info.csv: tabluar data describing the single datasets, their use in test or training and appropriate citation to the source publication wherre the data was first published.

This repository contains the data related to the publication

Steinn Ýmir Ágústsson, Mohammad Ahsanul Haque, Thi Tam Truong, Marco Bianchi, Nikita Klyuchnikov, Davide Mottin, Panagiotis Karras, Philip Hofmann; An autoencoder for compressing angle-resolved photoemission spectroscopy data. Mach. Learn.: Sci. Technol. 6 015019 (2025) DOI: 10.1088/2632-2153/ada8f2

Please cite the paper above in case of re-use of these data in a scientific publication.