Sparse Methods for Vector Embeddings of TPC Data

TL;DR

The paper addresses representation learning for Time Projection Chamber data using sparse convolutional networks to obtain transferable event embeddings across detectors. It leverages a sparse ResNet14 backbone implemented via the Minkowski Engine to process raw pad-level TPC hits as sparse tensors, trained on a simple proton–alpha classification task within GADGET II. The resulting embeddings transfer to the AT-TPC and show meaningful structure in linear probes and PCA, with pretraining enhancing both in-domain and cross-domain performance. These findings demonstrate the viability of sparse tensor methods for cross-detector TPC representation learning and point toward developing cross-domain TPC foundation models.

Abstract

Time Projection Chambers (TPCs) are versatile detectors that reconstruct charged-particle tracks in an ionizing medium, enabling sensitive measurements across a wide range of nuclear physics experiments. We explore sparse convolutional networks for representation learning on TPC data, finding that a sparse ResNet architecture, even with randomly set weights, provides useful structured vector embeddings of events. Pre-training this architecture on a simple physics-motivated binary classification task further improves the embedding quality. Using data from the GAseous Detector with GErmanium Tagging (GADGET) II TPC, a detector optimized for measuring low-energy $β$-delayed particle decays, we represent raw pad-level signals as sparse tensors, train Minkowski Engine ResNet models, and probe the resulting event-level embeddings which reveal rich event structure. As a cross-detector test, we embed data from the Active-Target TPC (AT-TPC) -- a detector designed for nuclear reaction studies in inverse kinematics -- using the same encoder. We find that even an untrained sparse ResNet model provides useful embeddings of AT-TPC data, and we observe improvements when the model is trained on GADGET data. Together, these results highlight the potential of sparse convolutional techniques as a general tool for representation learning in diverse TPC experiments.

Figures (3)

• Figure 1: Left: an example of a 4-track AT-TPC event. Right: an example of a 1600 keV proton event from the GADGET II TPC.
• Figure 2: Left: Range vs energy plot for FRIB experiment 21072 (1 hr run). Distinct proton and alpha bands are visible; gating cuts on these bands provided training labels for models, while events in the overlap region were excluded. Middle: PCA of embeddings from ResNet$_{rand}$ . Right: PCA of embeddings from ResNet$_{train}$ . In PCA plots: 800 keV $p$, 1600 keV $p$, and 2 MeV $\alpha$; brown, light blue, and dark blue, respectively.
• Figure 3: Visualization of AT-TPC ResNet embeddings. Plot a) shows the first two principal components (PCA1 and PCA2 respectively) of embeddings from ResNet$_{rand}$. Plot b) depicts PCA2 vs. PCA1 from ResNet$_{train}$. Plots c) and d) present the distribution of latent embeddings for PCA1 and PCA2, respectively.