Influencer Loss: End-to-end Geometric Representation Learning for Track Reconstruction

被引:0
|
作者
Murnane, Daniel [1 ]
机构
[1] Lawrence Berkeley Natl Lab, Berkeley, CA 94720 USA
来源
26TH INTERNATIONAL CONFERENCE ON COMPUTING IN HIGH ENERGY AND NUCLEAR PHYSICS, CHEP 2023 | 2024年 / 295卷
关键词
D O I
10.1051/epjconf/202429509016
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
Significant progress has been made in applying graph neural networks (GNNs) and other geometric ML ideas to the track reconstruction problem. State-of-the-art results are obtained using approaches such as the Exatrkx pipeline, which currently applies separate edge construction, classification and segmentation stages. One can also treat the problem as an object condensation task, and cluster hits into tracks in a single stage, such as in the GravNet architecture. However, condensation with such an architecture may still require non-differentiable operations, and arbitrary post-processing. In this work, I extend the ideas of geometric attention to the task of fully geometric (and therefore fully differentiable) end-to-end track reconstruction in a single step. To realize this goal, I introduce a novel condensation loss function called Influencer Loss, which allows an embedded representation of tracks to be learned in tandem with the most representative hit(s) in each track. This loss has global optima that formally match the task of track reconstruction, namely smooth condensation of tracks to a single point, and I demonstrate this empirically on the TrackML dataset. The model not only significantly outperforms the physics performance of the baseline model, it is up to an order of magnitude faster in inference.
引用
收藏
页数:8
相关论文
共 50 条
  • [41] End-to-End Surface Reconstruction for Touching Trajectories
    Liu, Jiarui
    Zhang, Yuanpei
    Zou, Zhuojun
    Hao, Jie
    COMPUTER VISION - ACCV 2022, PT VII, 2023, 13847 : 101 - 116
  • [42] Geometric Transformer for End-to-End Molecule Properties Prediction
    Choukroun, Yoni
    Wolf, Lior
    PROCEEDINGS OF THE THIRTY-FIRST INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, IJCAI 2022, 2022, : 2895 - 2901
  • [43] STATISTICAL LEARNING FOR END-TO-END SIMULATIONS
    Vicent, J.
    Verrelst, J.
    Rivera-Caicedo, J. P.
    Sabater, N.
    Munoz-Mari, J.
    Camps-Valls, G.
    Moreno, J.
    IGARSS 2018 - 2018 IEEE INTERNATIONAL GEOSCIENCE AND REMOTE SENSING SYMPOSIUM, 2018, : 1699 - 1702
  • [44] END-TO-END LEARNING FOR MUSIC AUDIO
    Dieleman, Sander
    Schrauwen, Benjamin
    2014 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), 2014,
  • [45] End-to-end Learning for Graph Decomposition
    Song, Jie
    Andres, Bjoern
    Black, Michael J.
    Hilliges, Otmar
    Tang, Siyu
    2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, : 10092 - 10101
  • [46] Amharic OCR: An End-to-End Learning
    Belay, Birhanu
    Habtegebrial, Tewodros
    Meshesha, Million
    Liwicki, Marcus
    Belay, Gebeyehu
    Stricker, Didier
    APPLIED SCIENCES-BASEL, 2020, 10 (03):
  • [47] The Predictron: End-To-End Learning and Planning
    Silver, David
    van Hasselt, Hado
    Hessel, Matteo
    Schaul, Tom
    Guez, Arthur
    Harley, Tim
    Dulac-Arnold, Gabriel
    Reichert, David
    Rabinowitz, Neil
    Barret, Andre
    Degris, Thomas
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 70, 2017, 70
  • [48] End-to-End Photo-Sketch Generation via Fully Convolutional Representation Learning
    Zhang, Liliang
    Lin, Liang
    Wu, Xian
    Ding, Shengyong
    Zhang, Lei
    ICMR'15: PROCEEDINGS OF THE 2015 ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA RETRIEVAL, 2015, : 627 - 634
  • [49] Toward End-to-end Prediction of Future Wellbeing using Deep Sensor Representation Learning
    Li, Boning
    Yu, Han
    Sano, Akane
    2019 8TH INTERNATIONAL CONFERENCE ON AFFECTIVE COMPUTING AND INTELLIGENT INTERACTION WORKSHOPS AND DEMOS (ACIIW), 2019, : 253 - 257
  • [50] END-TO-END LEARNING OF COMPRESSIBLE FEATURES
    Singh, Saurabh
    Abu-El-Haija, Sami
    Johnston, Nick
    Balle, Johannes
    Shrivastava, Abhinav
    Toderici, George
    2020 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2020, : 3349 - 3353