{
  "id": 63316,
  "title": "Pretty pictures: What's up with B at high +z?",
  "url": "/competitions/trackml-particle-identification/discussion/63316",
  "author_name": "",
  "post_date": "2018-08-14T19:32:14.879773400Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Here are some nice pictures I wanted to share. They show the systematic deviations of intersection points from the ideal helix predictions in azimuthal (qduphi_p) and radial (qdur2_p) direction for caps and in azimuthal and z-direction (qdz_p) for cylinders. To get the actual deviation for a particle, you need to multiply by the charge of the particle and divide roughly by the curvature radius of the helix.</p>\n\n<p>Some things to notice:</p>\n\n<ul>\n<li>The overall oblique gradient (e.g. in qdur2_p) indicates a slight misalignment of the B field with the z-axis, I think.</li>\n<li>The general azimuthal shift at the end caps indicates a weakening of the B_z component with increasing magnitude |z|.</li>\n<li>The B-field gets really crazy at the end cap at positive z (cap_id = 23 in my internal nomenclature).</li>\n</ul>\n\n<p>Note: For each cap/cylinder my code created up to three pictures due to grouping helices by the coordinates of the previous layer intersection (z for caps, r2 for cylinders).</p>\n\n<p>Note: The larger images have transparent parts. If your browser shows a lot of black, save the downloaded image and open in a decent image viewer.</p>\n\n<p>Note: For the z-deviation in cylinders it's potentially interesting to look at both the charge-dependent and charge-independent one. I ended up using only the charge dependent one as it helped my score a bit more. Ideally both could be learned and used, but that did not work out so well for me.</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10077/dp_qduphi_p_caps_detail.png\" alt=\"helix deviations at the last cap at positive z, azimuthal\">\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10076/dp_qdur2_p_caps_detail.png\" alt=\"helix deviations at the last cap at positive z, radial\"></p>\n\n<p>[I hope I get the image attachment right.]</p>",
  "messages": [
    {
      "id": "370415",
      "postDate": "08/14/2018 19:32:14",
      "content": "<p>Here are some nice pictures I wanted to share. They show the systematic deviations of intersection points from the ideal helix predictions in azimuthal (qduphi_p) and radial (qdur2_p) direction for caps and in azimuthal and z-direction (qdz_p) for cylinders. To get the actual deviation for a particle, you need to multiply by the charge of the particle and divide roughly by the curvature radius of the helix.</p>\n\n<p>Some things to notice:</p>\n\n<ul>\n<li>The overall oblique gradient (e.g. in qdur2_p) indicates a slight misalignment of the B field with the z-axis, I think.</li>\n<li>The general azimuthal shift at the end caps indicates a weakening of the B_z component with increasing magnitude |z|.</li>\n<li>The B-field gets really crazy at the end cap at positive z (cap_id = 23 in my internal nomenclature).</li>\n</ul>\n\n<p>Note: For each cap/cylinder my code created up to three pictures due to grouping helices by the coordinates of the previous layer intersection (z for caps, r2 for cylinders).</p>\n\n<p>Note: The larger images have transparent parts. If your browser shows a lot of black, save the downloaded image and open in a decent image viewer.</p>\n\n<p>Note: For the z-deviation in cylinders it's potentially interesting to look at both the charge-dependent and charge-independent one. I ended up using only the charge dependent one as it helped my score a bit more. Ideally both could be learned and used, but that did not work out so well for me.</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10077/dp_qduphi_p_caps_detail.png\" alt=\"helix deviations at the last cap at positive z, azimuthal\">\n<img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10076/dp_qdur2_p_caps_detail.png\" alt=\"helix deviations at the last cap at positive z, radial\"></p>\n\n<p>[I hope I get the image attachment right.]</p>",
      "rawMarkdown": "Here are some nice pictures I wanted to share. They show the systematic deviations of intersection points from the ideal helix predictions in azimuthal (qduphi_p) and radial (qdur2_p) direction for caps and in azimuthal and z-direction (qdz_p) for cylinders. To get the actual deviation for a particle, you need to multiply by the charge of the particle and divide roughly by the curvature radius of the helix.\n\nSome things to notice:\n\n - The overall oblique gradient (e.g. in qdur2_p) indicates a slight misalignment of the B field with the z-axis, I think.\n - The general azimuthal shift at the end caps indicates a weakening of the B_z component with increasing magnitude |z|.\n - The B-field gets really crazy at the end cap at positive z (cap_id = 23 in my internal nomenclature).\n\nNote: For each cap/cylinder my code created up to three pictures due to grouping helices by the coordinates of the previous layer intersection (z for caps, r2 for cylinders).\n\nNote: The larger images have transparent parts. If your browser shows a lot of black, save the downloaded image and open in a decent image viewer.\n\nNote: For the z-deviation in cylinders it's potentially interesting to look at both the charge-dependent and charge-independent one. I ended up using only the charge dependent one as it helped my score a bit more. Ideally both could be learned and used, but that did not work out so well for me.\n\n![helix deviations at the last cap at positive z, azimuthal][1]\n![helix deviations at the last cap at positive z, radial][2]\n\n[I hope I get the image attachment right.]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10077/dp_qduphi_p_caps_detail.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10076/dp_qdur2_p_caps_detail.png",
      "votes": null
    },
    {
      "id": "370641",
      "postDate": "08/15/2018 07:33:37",
      "content": "<p>Nice pictures.  You make me want to share some of mine too ;)</p>",
      "rawMarkdown": "Nice pictures.  You make me want to share some of mine too ;)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 370641,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "08/15/2018 07:33:37",
      "content": "<p>Nice pictures.  You make me want to share some of mine too ;)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "370415": "Here are some nice pictures I wanted to share. They show the systematic deviations of intersection points from the ideal helix predictions in azimuthal (qduphi_p) and radial (qdur2_p) direction for caps and in azimuthal and z-direction (qdz_p) for cylinders. To get the actual deviation for a particle, you need to multiply by the charge of the particle and divide roughly by the curvature radius of the helix.\n\nSome things to notice:\n\n - The overall oblique gradient (e.g. in qdur2_p) indicates a slight misalignment of the B field with the z-axis, I think.\n - The general azimuthal shift at the end caps indicates a weakening of the B_z component with increasing magnitude |z|.\n - The B-field gets really crazy at the end cap at positive z (cap_id = 23 in my internal nomenclature).\n\nNote: For each cap/cylinder my code created up to three pictures due to grouping helices by the coordinates of the previous layer intersection (z for caps, r2 for cylinders).\n\nNote: The larger images have transparent parts. If your browser shows a lot of black, save the downloaded image and open in a decent image viewer.\n\nNote: For the z-deviation in cylinders it's potentially interesting to look at both the charge-dependent and charge-independent one. I ended up using only the charge dependent one as it helped my score a bit more. Ideally both could be learned and used, but that did not work out so well for me.\n\n![helix deviations at the last cap at positive z, azimuthal][1]\n![helix deviations at the last cap at positive z, radial][2]\n\n[I hope I get the image attachment right.]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10077/dp_qduphi_p_caps_detail.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/370415/10076/dp_qdur2_p_caps_detail.png",
    "370641": "Nice pictures.  You make me want to share some of mine too ;)"
  },
  "source": "meta"
}