Convergent approaches to AI Explainability for HEP muonic particles pattern recognition Dataset

Dataset associated to the publication "Convergent approaches to AI Explainability for HEP muonic particles pattern recognition", Leandro Maglianella, Lorenzo Nicoletti, Stefano Giagu*, Christian Napoli, and Simone Scardapane, submitted to Computing and Software for Big Science.

*corresponding author: stefano.giagu [AT] uniroma1.it

Description:

provided as a compressed zip file. Contains 7 numpy .npy files:

• train_images_with_noise.npy: numpy array containing 850003 "images" of muonic tracks with detector noise (shape (850003, 9, 384)). Each image contains 1 muonic track.
• train_images_without_noise.npy: numpy array containing 850003 "images" of muonic tracks w/o detector noise (shape (850003, 9, 384)). Each image contains 1 muonic track.
• train_labels.npy: labels associated to each image (shape (850003, 5)), corresponding to  (pT, eta, phi, 0, nhits) of the muonic track, with pT: transverse momentum, eta: pseudo-rapidity,  phi: azimuthal angle, and nhits: the number of pixels turned on by the muon
• test_images_with_noise.npy: same as above for a 94445 images test set
• test_images_without_noise.npy: same as above for a 94445 images test set
• test_labels.npy: same as above for a 94445 images test set
• images_only_noise.npy: numpy array containing 944448 "images" w/o muons, containing detector noise only (shape (944448, 9, 384))