{
  "id": 56668,
  "title": "Data questions (issues?)",
  "url": "/competitions/trackml-particle-identification/discussion/56668",
  "author_name": "",
  "post_date": "2018-05-13T06:17:56.065097Z",
  "votes": 4,
  "comment_count": 24,
  "views": 0,
  "content": "<p>For particle 117109052193898496 there is a strange behaviour of momentum.</p>\n\n<p>Component Pz is rapidly decreases: -2.81045, -2.6336, -1.81182, -1.55942, -1.57022.</p>\n\n<p>Tangential momentum Pxy = sqrt(Px^2+Py^2) increases: 0.146478, 0.991588, 2.15149, 2.33823, 2.33013.</p>\n\n<p>How is it possible physically?</p>",
  "messages": [
    {
      "id": "328008",
      "postDate": "05/13/2018 06:17:56",
      "content": "<p>For particle 117109052193898496 there is a strange behaviour of momentum.</p>\n\n<p>Component Pz is rapidly decreases: -2.81045, -2.6336, -1.81182, -1.55942, -1.57022.</p>\n\n<p>Tangential momentum Pxy = sqrt(Px^2+Py^2) increases: 0.146478, 0.991588, 2.15149, 2.33823, 2.33013.</p>\n\n<p>How is it possible physically?</p>",
      "rawMarkdown": "For particle 117109052193898496 there is a strange behaviour of momentum.\n\nComponent Pz is rapidly decreases: -2.81045, -2.6336, -1.81182, -1.55942, -1.57022.\n\nTangential momentum Pxy = sqrt(Px^2+Py^2) increases: 0.146478, 0.991588, 2.15149, 2.33823, 2.33013.\n\nHow is it possible physically?",
      "votes": null
    },
    {
      "id": "328014",
      "postDate": "05/13/2018 06:41:58",
      "content": "<p>For 4532667966029824 the coordinate Z changes nonmonotonically: -0.0778789, -0.562072, -0.569675, -1.03353, -1.36867, -1.79654, -1.80895, -2.25398, -2.32808, -1.43974, 0.61262, 3.27742, 6.73515.</p>",
      "rawMarkdown": "For 4532667966029824 the coordinate Z changes nonmonotonically: -0.0778789, -0.562072, -0.569675, -1.03353, -1.36867, -1.79654, -1.80895, -2.25398, -2.32808, -1.43974, 0.61262, 3.27742, 6.73515.",
      "votes": null
    },
    {
      "id": "328019",
      "postDate": "05/13/2018 06:48:04",
      "content": "<p>conservation of (total) momentum? </p>",
      "rawMarkdown": "conservation of (total) momentum?",
      "votes": null
    },
    {
      "id": "328023",
      "postDate": "05/13/2018 07:04:25",
      "content": "<p>Particle 117110014266572800. Momentum plot.\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/328023/9410/f1.png\" alt=\"117110014266572800 Momentum \"></p>\n\n<p>?1 - What may cause such fluctuations?</p>\n\n<p>?2 - Magnetic field changes direction (to opposite) ?</p>",
      "rawMarkdown": "Particle 117110014266572800. Momentum plot.\n![117110014266572800 Momentum ][1]\n\n?1 - What may cause such fluctuations?\n\n?2 - Magnetic field changes direction (to opposite) ?\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/328023/9410/f1.png",
      "votes": null
    },
    {
      "id": "328029",
      "postDate": "05/13/2018 07:26:07",
      "content": "<p>Total (absolute value) momentum looks OK. But I can't guess what may change  Z component of momentum. Magnetic field cannot change collinear momentum.</p>",
      "rawMarkdown": "Total (absolute value) momentum looks OK. But I can't guess what may change  Z component of momentum. Magnetic field cannot change collinear momentum.",
      "votes": null
    },
    {
      "id": "328033",
      "postDate": "05/13/2018 07:39:05",
      "content": "<p>There's a loss of energy as the particle interacts with the materials in the detector</p>",
      "rawMarkdown": "There's a loss of energy as the particle interacts with the materials in the detector",
      "votes": null
    },
    {
      "id": "328043",
      "postDate": "05/13/2018 08:09:10",
      "content": "<blockquote>\n  <p><strong>aeevrubin wrote</strong></p>\n  \n  <blockquote>\n    <p>There's a loss of energy as the particle interacts with the materials in the detector</p>\n  </blockquote>\n</blockquote>\n\n<p>Yes, but this has not so strong effect. </p>\n\n<p>And how such interaction with the materials can (systematically!) increase Z component of particle velocity?</p>",
      "rawMarkdown": "&gt; **aeevrubin wrote**\n&gt; \n&gt; &gt; There's a loss of energy as the particle interacts with the materials in the detector\n\nYes, but this has not so strong effect. \n\nAnd how such interaction with the materials can (systematically!) increase Z component of particle velocity?",
      "votes": null
    },
    {
      "id": "328044",
      "postDate": "05/13/2018 08:12:00",
      "content": "<p>The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? </p>",
      "rawMarkdown": "The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right?",
      "votes": null
    },
    {
      "id": "328047",
      "postDate": "05/13/2018 08:18:48",
      "content": "<blockquote>\n  <p><strong>aeevrubin wrote</strong></p>\n  \n  <blockquote>\n    <p>The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? </p>\n  </blockquote>\n</blockquote>\n\n<p>The \"truth\" files contain tpx, tpy tpz for all hits. We can follow momentum changes.</p>",
      "rawMarkdown": "&gt; **aeevrubin wrote**\n&gt; \n&gt; &gt; The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? \n\nThe \"truth\" files contain tpx, tpy tpz for all hits. We can follow momentum changes.",
      "votes": null
    },
    {
      "id": "328087",
      "postDate": "05/13/2018 10:07:47",
      "content": "<p><strong>Perhaps</strong> it has something to do with the fact that we are using simulated data. Of course, this is nothing more than an assumption.</p>",
      "rawMarkdown": "**Perhaps** it has something to do with the fact that we are using simulated data. Of course, this is nothing more than an assumption.",
      "votes": null
    },
    {
      "id": "328089",
      "postDate": "05/13/2018 10:11:13",
      "content": "<p>In a perfect scenario with a perfect detector that doesn't interact, the px~sin t and py~cos t, so px vs py would be a circle. Since the particle is losing energy it can never return back to it's initial momentum.\nEq 2.31, page 42, <a href=\"https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf\">https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf</a></p>\n\n<p>Your plot seems consistent with that given the detector is very imperfect and there are thousands of other particles whizzing through it, I guess</p>",
      "rawMarkdown": "In a perfect scenario with a perfect detector that doesn't interact, the px~sin t and py~cos t, so px vs py would be a circle. Since the particle is losing energy it can never return back to it's initial momentum.\nEq 2.31, page 42, https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf\n\nYour plot seems consistent with that given the detector is very imperfect and there are thousands of other particles whizzing through it, I guess",
      "votes": null
    },
    {
      "id": "328095",
      "postDate": "05/13/2018 10:32:19",
      "content": "<p>Can you tell me what event this is ? </p>",
      "rawMarkdown": "Can you tell me what event this is ?",
      "votes": null
    },
    {
      "id": "328097",
      "postDate": "05/13/2018 10:38:06",
      "content": "<p>I will look into this, change of the Pz component is indeed possible, because the magnetic field is not (0,0,Bz) throughout the detector but rather a realistic one which you get from a limited length solenoid coil. </p>",
      "rawMarkdown": "I will look into this, change of the Pz component is indeed possible, because the magnetic field is not (0,0,Bz) throughout the detector but rather a realistic one which you get from a limited length solenoid coil.",
      "votes": null
    },
    {
      "id": "328109",
      "postDate": "05/13/2018 11:07:51",
      "content": "<blockquote>\n  <p><strong>Andreas Salzburger wrote</strong></p>\n  \n  <blockquote>\n    <p>Can you tell me what event this is ? </p>\n  </blockquote>\n</blockquote>\n\n<p>event000001000-truth.csv</p>",
      "rawMarkdown": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; &gt; Can you tell me what event this is ? \n\nevent000001000-truth.csv",
      "votes": null
    },
    {
      "id": "328181",
      "postDate": "05/13/2018 15:28:09",
      "content": "<p>Can you confirm that not even volumeId, layerId, moduleId, ch0, ch1 is unique?</p>\n\n<p>Event 1000, hitIds 8533 &amp; 8535 have volumeId, layerId, moduleId, ch0, ch1 = 7, 10, 23, 252, 105 ?</p>",
      "rawMarkdown": "Can you confirm that not even volumeId, layerId, moduleId, ch0, ch1 is unique?\n\nEvent 1000, hitIds 8533 &amp; 8535 have volumeId, layerId, moduleId, ch0, ch1 = 7, 10, 23, 252, 105 ?",
      "votes": null
    },
    {
      "id": "328196",
      "postDate": "05/13/2018 16:09:05",
      "content": "<p>Track 4550397591027712 from event 1000 also looks strange:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/328196/9417/f2.png\" alt=\"XY plot\"></p>",
      "rawMarkdown": "Track 4550397591027712 from event 1000 also looks strange:\n![XY plot][1]\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/328196/9417/f2.png",
      "votes": null
    },
    {
      "id": "328490",
      "postDate": "05/14/2018 13:16:05",
      "content": "<p>For your information: the magnetic field is not completely aligned with the z-axis (like in reality, when you build a detector, your detector coil might be a bit tilted with respect to the detector), I am just validating that what you see is exactly this change.</p>",
      "rawMarkdown": "For your information: the magnetic field is not completely aligned with the z-axis (like in reality, when you build a detector, your detector coil might be a bit tilted with respect to the detector), I am just validating that what you see is exactly this change.",
      "votes": null
    },
    {
      "id": "328491",
      "postDate": "05/14/2018 13:17:10",
      "content": "<p>I am looking into this, and come back to you. \nJust not that in the outer detectors - due to the long strips - you can not expect the trajectory to be fully smooth in visualization.</p>",
      "rawMarkdown": "I am looking into this, and come back to you. \nJust not that in the outer detectors - due to the long strips - you can not expect the trajectory to be fully smooth in visualization.",
      "votes": null
    },
    {
      "id": "328581",
      "postDate": "05/14/2018 16:35:08",
      "content": "<blockquote>\n  <p><strong>Andreas Salzburger wrote</strong></p>\n  \n  <p>.. you can not expect the trajectory to be fully smooth in visualization.</p>\n</blockquote>\n\n<p>Why not? I plot the trajectory using the ground truth data. Not measured data. How the <strong>true</strong> trajectory may be not smooth?</p>",
      "rawMarkdown": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; .. you can not expect the trajectory to be fully smooth in visualization.\n\nWhy not? I plot the trajectory using the ground truth data. Not measured data. How the **true** trajectory may be not smooth?",
      "votes": null
    },
    {
      "id": "328586",
      "postDate": "05/14/2018 16:45:42",
      "content": "<p>&gt; <strong>Andreas Salzburger wrote</strong>\n&gt; \n&gt; &gt; For your information: the magnetic field is not completely aligned with the z-axis..</p>\n\n<p>I suppose that \"not completely aligned\" does not mean \"almost orthogonal\". What I see for 117109052193898496 may be explaiyned if the magnetic field is almost orthogonal to Z.</p>",
      "rawMarkdown": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; &gt; For your information: the magnetic field is not completely aligned with the z-axis..\n\nI suppose that \"not completely aligned\" does not mean \"almost orthogonal\". What I see for 117109052193898496 may be explaiyned if the magnetic field is almost orthogonal to Z.",
      "votes": null
    },
    {
      "id": "328640",
      "postDate": "05/14/2018 19:27:59",
      "content": "<p>The <em>true</em> trajectory is smooth (except for scattering effects, of course which give slight kinks), indeed, but you do not have points between the single measurements, hence creating a smooth image is not easy (one can do some Bezier interpolation or some other sort).</p>\n\n<p>As I said - this specific case I am looking at already, I will come back with my finding.</p>",
      "rawMarkdown": "The *true* trajectory is smooth (except for scattering effects, of course which give slight kinks), indeed, but you do not have points between the single measurements, hence creating a smooth image is not easy (one can do some Bezier interpolation or some other sort).\n\nAs I said - this specific case I am looking at already, I will come back with my finding.",
      "votes": null
    },
    {
      "id": "330598",
      "postDate": "05/19/2018 08:05:12",
      "content": "<p>Some randomly choosen tracks in momentum space:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9452/f3.png\" alt=\"Px and Py for some tracks\"></p>\n\n<p>Typically they are arcs. But some of them looks like the charge of particles change their sign (to opposite).</p>\n\n<p>Several tracks are quite different (4550397591027712, event 1000):\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9453/f4.png\" alt=\"Track 4550397591027712\"></p>",
      "rawMarkdown": "Some randomly choosen tracks in momentum space:\n![Px and Py for some tracks][1]\n\nTypically they are arcs. But some of them looks like the charge of particles change their sign (to opposite).\n\nSeveral tracks are quite different (4550397591027712, event 1000):\n![Track 4550397591027712][2]\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9452/f3.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9453/f4.png",
      "votes": null
    },
    {
      "id": "330607",
      "postDate": "05/19/2018 08:51:11",
      "content": "<p>Hough transform plot:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330607/9454/f5.png\" alt=\"Rxy on Z\">\nSome outliers are marked (colored).\nGreen track looks like the particle leaves collider and comes back.</p>",
      "rawMarkdown": "Hough transform plot:\n![Rxy on Z][1]\nSome outliers are marked (colored).\nGreen track looks like the particle leaves collider and comes back.\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/330607/9454/f5.png",
      "votes": null
    },
    {
      "id": "330684",
      "postDate": "05/19/2018 12:36:01",
      "content": "<p>Thanks for looking into that - I am trying to resimulate this exact one, stay tuned ... </p>",
      "rawMarkdown": "Thanks for looking into that - I am trying to resimulate this exact one, stay tuned ...",
      "votes": null
    },
    {
      "id": "335361",
      "postDate": "05/29/2018 16:32:54",
      "content": "<p>Hi, many thanks for looking into that.</p>\n\n<p>I re-simulated some of those awkward particles, and I can confirm that in very rare cases the navigation in the simulation goes wrong. I am currently quantifying the effect (it's at a few in 10000), and prepare a notebook with explanations and potentially a helper function to flag those.</p>\n\n<p>We will also test that those will not affect the score, of course.</p>",
      "rawMarkdown": "Hi, many thanks for looking into that.\n\nI re-simulated some of those awkward particles, and I can confirm that in very rare cases the navigation in the simulation goes wrong. I am currently quantifying the effect (it's at a few in 10000), and prepare a notebook with explanations and potentially a helper function to flag those.\n\nWe will also test that those will not affect the score, of course.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 328014,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "05/13/2018 06:41:58",
      "content": "<p>For 4532667966029824 the coordinate Z changes nonmonotonically: -0.0778789, -0.562072, -0.569675, -1.03353, -1.36867, -1.79654, -1.80895, -2.25398, -2.32808, -1.43974, 0.61262, 3.27742, 6.73515.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 328019,
      "author_name": "aeevrubin",
      "author_url": "",
      "post_date": "05/13/2018 06:48:04",
      "content": "<p>conservation of (total) momentum? </p>",
      "votes": null,
      "replies": [
        {
          "id": 328029,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/13/2018 07:26:07",
          "content": "<p>Total (absolute value) momentum looks OK. But I can't guess what may change  Z component of momentum. Magnetic field cannot change collinear momentum.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328033,
          "author_name": "aeevrubin",
          "author_url": "",
          "post_date": "05/13/2018 07:39:05",
          "content": "<p>There's a loss of energy as the particle interacts with the materials in the detector</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328043,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/13/2018 08:09:10",
          "content": "<blockquote>\n  <p><strong>aeevrubin wrote</strong></p>\n  \n  <blockquote>\n    <p>There's a loss of energy as the particle interacts with the materials in the detector</p>\n  </blockquote>\n</blockquote>\n\n<p>Yes, but this has not so strong effect. </p>\n\n<p>And how such interaction with the materials can (systematically!) increase Z component of particle velocity?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 328023,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "05/13/2018 07:04:25",
      "content": "<p>Particle 117110014266572800. Momentum plot.\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/328023/9410/f1.png\" alt=\"117110014266572800 Momentum \"></p>\n\n<p>?1 - What may cause such fluctuations?</p>\n\n<p>?2 - Magnetic field changes direction (to opposite) ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 328087,
          "author_name": "konchar",
          "author_url": "",
          "post_date": "05/13/2018 10:07:47",
          "content": "<p><strong>Perhaps</strong> it has something to do with the fact that we are using simulated data. Of course, this is nothing more than an assumption.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328089,
          "author_name": "aeevrubin",
          "author_url": "",
          "post_date": "05/13/2018 10:11:13",
          "content": "<p>In a perfect scenario with a perfect detector that doesn't interact, the px~sin t and py~cos t, so px vs py would be a circle. Since the particle is losing energy it can never return back to it's initial momentum.\nEq 2.31, page 42, <a href=\"https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf\">https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf</a></p>\n\n<p>Your plot seems consistent with that given the detector is very imperfect and there are thousands of other particles whizzing through it, I guess</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328095,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/13/2018 10:32:19",
          "content": "<p>Can you tell me what event this is ? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328097,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/13/2018 10:38:06",
          "content": "<p>I will look into this, change of the Pz component is indeed possible, because the magnetic field is not (0,0,Bz) throughout the detector but rather a realistic one which you get from a limited length solenoid coil. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328109,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/13/2018 11:07:51",
          "content": "<blockquote>\n  <p><strong>Andreas Salzburger wrote</strong></p>\n  \n  <blockquote>\n    <p>Can you tell me what event this is ? </p>\n  </blockquote>\n</blockquote>\n\n<p>event000001000-truth.csv</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328490,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/14/2018 13:16:05",
          "content": "<p>For your information: the magnetic field is not completely aligned with the z-axis (like in reality, when you build a detector, your detector coil might be a bit tilted with respect to the detector), I am just validating that what you see is exactly this change.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328586,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/14/2018 16:45:42",
          "content": "<p>&gt; <strong>Andreas Salzburger wrote</strong>\n&gt; \n&gt; &gt; For your information: the magnetic field is not completely aligned with the z-axis..</p>\n\n<p>I suppose that \"not completely aligned\" does not mean \"almost orthogonal\". What I see for 117109052193898496 may be explaiyned if the magnetic field is almost orthogonal to Z.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 328044,
      "author_name": "aeevrubin",
      "author_url": "",
      "post_date": "05/13/2018 08:12:00",
      "content": "<p>The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? </p>",
      "votes": null,
      "replies": [
        {
          "id": 328047,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/13/2018 08:18:48",
          "content": "<blockquote>\n  <p><strong>aeevrubin wrote</strong></p>\n  \n  <blockquote>\n    <p>The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? </p>\n  </blockquote>\n</blockquote>\n\n<p>The \"truth\" files contain tpx, tpy tpz for all hits. We can follow momentum changes.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 328181,
      "author_name": "glimmung",
      "author_url": "",
      "post_date": "05/13/2018 15:28:09",
      "content": "<p>Can you confirm that not even volumeId, layerId, moduleId, ch0, ch1 is unique?</p>\n\n<p>Event 1000, hitIds 8533 &amp; 8535 have volumeId, layerId, moduleId, ch0, ch1 = 7, 10, 23, 252, 105 ?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 328196,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "05/13/2018 16:09:05",
      "content": "<p>Track 4550397591027712 from event 1000 also looks strange:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/328196/9417/f2.png\" alt=\"XY plot\"></p>",
      "votes": null,
      "replies": [
        {
          "id": 328491,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/14/2018 13:17:10",
          "content": "<p>I am looking into this, and come back to you. \nJust not that in the outer detectors - due to the long strips - you can not expect the trajectory to be fully smooth in visualization.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328581,
          "author_name": "nedelko",
          "author_url": "",
          "post_date": "05/14/2018 16:35:08",
          "content": "<blockquote>\n  <p><strong>Andreas Salzburger wrote</strong></p>\n  \n  <p>.. you can not expect the trajectory to be fully smooth in visualization.</p>\n</blockquote>\n\n<p>Why not? I plot the trajectory using the ground truth data. Not measured data. How the <strong>true</strong> trajectory may be not smooth?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 328640,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/14/2018 19:27:59",
          "content": "<p>The <em>true</em> trajectory is smooth (except for scattering effects, of course which give slight kinks), indeed, but you do not have points between the single measurements, hence creating a smooth image is not easy (one can do some Bezier interpolation or some other sort).</p>\n\n<p>As I said - this specific case I am looking at already, I will come back with my finding.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 330598,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "05/19/2018 08:05:12",
      "content": "<p>Some randomly choosen tracks in momentum space:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9452/f3.png\" alt=\"Px and Py for some tracks\"></p>\n\n<p>Typically they are arcs. But some of them looks like the charge of particles change their sign (to opposite).</p>\n\n<p>Several tracks are quite different (4550397591027712, event 1000):\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9453/f4.png\" alt=\"Track 4550397591027712\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 330607,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "05/19/2018 08:51:11",
      "content": "<p>Hough transform plot:\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/330607/9454/f5.png\" alt=\"Rxy on Z\">\nSome outliers are marked (colored).\nGreen track looks like the particle leaves collider and comes back.</p>",
      "votes": null,
      "replies": [
        {
          "id": 330684,
          "author_name": "asalzburger",
          "author_url": "",
          "post_date": "05/19/2018 12:36:01",
          "content": "<p>Thanks for looking into that - I am trying to resimulate this exact one, stay tuned ... </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 335361,
      "author_name": "asalzburger",
      "author_url": "",
      "post_date": "05/29/2018 16:32:54",
      "content": "<p>Hi, many thanks for looking into that.</p>\n\n<p>I re-simulated some of those awkward particles, and I can confirm that in very rare cases the navigation in the simulation goes wrong. I am currently quantifying the effect (it's at a few in 10000), and prepare a notebook with explanations and potentially a helper function to flag those.</p>\n\n<p>We will also test that those will not affect the score, of course.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "328008": "For particle 117109052193898496 there is a strange behaviour of momentum.\n\nComponent Pz is rapidly decreases: -2.81045, -2.6336, -1.81182, -1.55942, -1.57022.\n\nTangential momentum Pxy = sqrt(Px^2+Py^2) increases: 0.146478, 0.991588, 2.15149, 2.33823, 2.33013.\n\nHow is it possible physically?",
    "328014": "For 4532667966029824 the coordinate Z changes nonmonotonically: -0.0778789, -0.562072, -0.569675, -1.03353, -1.36867, -1.79654, -1.80895, -2.25398, -2.32808, -1.43974, 0.61262, 3.27742, 6.73515.",
    "328019": "conservation of (total) momentum?",
    "328023": "Particle 117110014266572800. Momentum plot.\n![117110014266572800 Momentum ][1]\n\n?1 - What may cause such fluctuations?\n\n?2 - Magnetic field changes direction (to opposite) ?\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/328023/9410/f1.png",
    "328029": "Total (absolute value) momentum looks OK. But I can't guess what may change  Z component of momentum. Magnetic field cannot change collinear momentum.",
    "328033": "There's a loss of energy as the particle interacts with the materials in the detector",
    "328043": "&gt; **aeevrubin wrote**\n&gt; \n&gt; &gt; There's a loss of energy as the particle interacts with the materials in the detector\n\nYes, but this has not so strong effect. \n\nAnd how such interaction with the materials can (systematically!) increase Z component of particle velocity?",
    "328044": "The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right?",
    "328047": "&gt; **aeevrubin wrote**\n&gt; \n&gt; &gt; The pdf says: \"px,py,pz: initial momentum (in GeV/c) at particle vertex\". A single particle track should only have one (px,py,pz) vector. Right? \n\nThe \"truth\" files contain tpx, tpy tpz for all hits. We can follow momentum changes.",
    "328087": "**Perhaps** it has something to do with the fact that we are using simulated data. Of course, this is nothing more than an assumption.",
    "328089": "In a perfect scenario with a perfect detector that doesn't interact, the px~sin t and py~cos t, so px vs py would be a circle. Since the particle is losing energy it can never return back to it's initial momentum.\nEq 2.31, page 42, https://ocw.mit.edu/courses/nuclear-engineering/22-105-electromagnetic-interactions-fall-2005/readings/chap2.pdf\n\nYour plot seems consistent with that given the detector is very imperfect and there are thousands of other particles whizzing through it, I guess",
    "328095": "Can you tell me what event this is ?",
    "328097": "I will look into this, change of the Pz component is indeed possible, because the magnetic field is not (0,0,Bz) throughout the detector but rather a realistic one which you get from a limited length solenoid coil.",
    "328109": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; &gt; Can you tell me what event this is ? \n\nevent000001000-truth.csv",
    "328181": "Can you confirm that not even volumeId, layerId, moduleId, ch0, ch1 is unique?\n\nEvent 1000, hitIds 8533 &amp; 8535 have volumeId, layerId, moduleId, ch0, ch1 = 7, 10, 23, 252, 105 ?",
    "328196": "Track 4550397591027712 from event 1000 also looks strange:\n![XY plot][1]\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/328196/9417/f2.png",
    "328490": "For your information: the magnetic field is not completely aligned with the z-axis (like in reality, when you build a detector, your detector coil might be a bit tilted with respect to the detector), I am just validating that what you see is exactly this change.",
    "328491": "I am looking into this, and come back to you. \nJust not that in the outer detectors - due to the long strips - you can not expect the trajectory to be fully smooth in visualization.",
    "328581": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; .. you can not expect the trajectory to be fully smooth in visualization.\n\nWhy not? I plot the trajectory using the ground truth data. Not measured data. How the **true** trajectory may be not smooth?",
    "328586": "&gt; **Andreas Salzburger wrote**\n&gt; \n&gt; &gt; For your information: the magnetic field is not completely aligned with the z-axis..\n\nI suppose that \"not completely aligned\" does not mean \"almost orthogonal\". What I see for 117109052193898496 may be explaiyned if the magnetic field is almost orthogonal to Z.",
    "328640": "The *true* trajectory is smooth (except for scattering effects, of course which give slight kinks), indeed, but you do not have points between the single measurements, hence creating a smooth image is not easy (one can do some Bezier interpolation or some other sort).\n\nAs I said - this specific case I am looking at already, I will come back with my finding.",
    "330598": "Some randomly choosen tracks in momentum space:\n![Px and Py for some tracks][1]\n\nTypically they are arcs. But some of them looks like the charge of particles change their sign (to opposite).\n\nSeveral tracks are quite different (4550397591027712, event 1000):\n![Track 4550397591027712][2]\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9452/f3.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/330598/9453/f4.png",
    "330607": "Hough transform plot:\n![Rxy on Z][1]\nSome outliers are marked (colored).\nGreen track looks like the particle leaves collider and comes back.\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/330607/9454/f5.png",
    "330684": "Thanks for looking into that - I am trying to resimulate this exact one, stay tuned ...",
    "335361": "Hi, many thanks for looking into that.\n\nI re-simulated some of those awkward particles, and I can confirm that in very rare cases the navigation in the simulation goes wrong. I am currently quantifying the effect (it's at a few in 10000), and prepare a notebook with explanations and potentially a helper function to flag those.\n\nWe will also test that those will not affect the score, of course."
  },
  "source": "meta"
}