{
  "id": 68144,
  "title": "template light curve",
  "url": "/competitions/PLAsTiCC-2018/discussion/68144",
  "author_name": "",
  "post_date": "2018-10-09T19:03:59.443304Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I know that the light curves are the result of difference imaging between two separate nights that is why there are negative flux. Is it possible to have the information of the template light curve that was used to make the substraction for each object? Indeed I am interesting in having the original light curve before the substraction to have only positive fluxes and eventually transform the fluxes in magnitudes.</p>\n\n<p>Thank you.</p>",
  "messages": [
    {
      "id": "401289",
      "postDate": "10/09/2018 19:03:59",
      "content": "<p>I know that the light curves are the result of difference imaging between two separate nights that is why there are negative flux. Is it possible to have the information of the template light curve that was used to make the substraction for each object? Indeed I am interesting in having the original light curve before the substraction to have only positive fluxes and eventually transform the fluxes in magnitudes.</p>\n\n<p>Thank you.</p>",
      "rawMarkdown": "I know that the light curves are the result of difference imaging between two separate nights that is why there are negative flux. Is it possible to have the information of the template light curve that was used to make the substraction for each object? Indeed I am interesting in having the original light curve before the substraction to have only positive fluxes and eventually transform the fluxes in magnitudes.\n\nThank you.",
      "votes": null
    },
    {
      "id": "401672",
      "postDate": "10/10/2018 13:21:06",
      "content": "<p>Hi Jerome,</p>\n\n<p>Template flux isn't included in the challenge dataset. </p>\n\n<p>That said... this information will be made available along with other metadata after the conclusion of the challenge (both here on Kaggle and for the \"science prize\" entries), and we anticipate the full dataset will prove useful for other studies. Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. </p>\n\n<p>So you'll absolutely be able to get what you are interested in, but not for a bit. </p>\n\n<p>-Gautham on behalf of the PLAsTiCC team</p>",
      "rawMarkdown": "Hi Jerome,\n\nTemplate flux isn't included in the challenge dataset. \n\nThat said... this information will be made available along with other metadata after the conclusion of the challenge (both here on Kaggle and for the \"science prize\" entries), and we anticipate the full dataset will prove useful for other studies. Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. \n\nSo you'll absolutely be able to get what you are interested in, but not for a bit. \n\n-Gautham on behalf of the PLAsTiCC team",
      "votes": null
    },
    {
      "id": "402741",
      "postDate": "10/12/2018 08:24:32",
      "content": "<blockquote>\n  <p>Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. </p>\n</blockquote>\n\n<p>@Gautham, Do not treat (real) fluxes as metadata - they are base data. As I well understand, that what you gave us is a relative flux, relative in the sense of subtraction of its historic value. In the case of cyclic objects, logarithms of such values make no sens, so we cannot use a slope of magnitude or colors (based on logarithms) succesfully used in astronomy for object classification. I am sure, the winners of the competition will find some new features, but the features will play the role of the actors of the third plan in \"normal\" classification based on backgroud corrected fluxes. It would be a great mistake to compare the results of this competition working on fluxes with this \"history\" correction with other algorithms working on fluxes with background correction.</p>",
      "rawMarkdown": "&gt;Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. \n\n@Gautham, Do not treat (real) fluxes as metadata - they are base data. As I well understand, that what you gave us is a relative flux, relative in the sense of subtraction of its historic value. In the case of cyclic objects, logarithms of such values make no sens, so we cannot use a slope of magnitude or colors (based on logarithms) succesfully used in astronomy for object classification. I am sure, the winners of the competition will find some new features, but the features will play the role of the actors of the third plan in \"normal\" classification based on backgroud corrected fluxes. It would be a great mistake to compare the results of this competition working on fluxes with this \"history\" correction with other algorithms working on fluxes with background correction.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 401672,
      "author_name": "gsnarayan",
      "author_url": "",
      "post_date": "10/10/2018 13:21:06",
      "content": "<p>Hi Jerome,</p>\n\n<p>Template flux isn't included in the challenge dataset. </p>\n\n<p>That said... this information will be made available along with other metadata after the conclusion of the challenge (both here on Kaggle and for the \"science prize\" entries), and we anticipate the full dataset will prove useful for other studies. Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. </p>\n\n<p>So you'll absolutely be able to get what you are interested in, but not for a bit. </p>\n\n<p>-Gautham on behalf of the PLAsTiCC team</p>",
      "votes": null,
      "replies": [
        {
          "id": 402741,
          "author_name": "sionek",
          "author_url": "",
          "post_date": "10/12/2018 08:24:32",
          "content": "<blockquote>\n  <p>Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. </p>\n</blockquote>\n\n<p>@Gautham, Do not treat (real) fluxes as metadata - they are base data. As I well understand, that what you gave us is a relative flux, relative in the sense of subtraction of its historic value. In the case of cyclic objects, logarithms of such values make no sens, so we cannot use a slope of magnitude or colors (based on logarithms) succesfully used in astronomy for object classification. I am sure, the winners of the competition will find some new features, but the features will play the role of the actors of the third plan in \"normal\" classification based on backgroud corrected fluxes. It would be a great mistake to compare the results of this competition working on fluxes with this \"history\" correction with other algorithms working on fluxes with background correction.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "401289": "I know that the light curves are the result of difference imaging between two separate nights that is why there are negative flux. Is it possible to have the information of the template light curve that was used to make the substraction for each object? Indeed I am interesting in having the original light curve before the substraction to have only positive fluxes and eventually transform the fluxes in magnitudes.\n\nThank you.",
    "401672": "Hi Jerome,\n\nTemplate flux isn't included in the challenge dataset. \n\nThat said... this information will be made available along with other metadata after the conclusion of the challenge (both here on Kaggle and for the \"science prize\" entries), and we anticipate the full dataset will prove useful for other studies. Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. \n\nSo you'll absolutely be able to get what you are interested in, but not for a bit. \n\n-Gautham on behalf of the PLAsTiCC team",
    "402741": "&gt;Many of the science prize metrics we're considering looks at the performance of the classification schemes vs this additional metadata so that we can quantify any subtle biases in the algorithms. \n\n@Gautham, Do not treat (real) fluxes as metadata - they are base data. As I well understand, that what you gave us is a relative flux, relative in the sense of subtraction of its historic value. In the case of cyclic objects, logarithms of such values make no sens, so we cannot use a slope of magnitude or colors (based on logarithms) succesfully used in astronomy for object classification. I am sure, the winners of the competition will find some new features, but the features will play the role of the actors of the third plan in \"normal\" classification based on backgroud corrected fluxes. It would be a great mistake to compare the results of this competition working on fluxes with this \"history\" correction with other algorithms working on fluxes with background correction."
  },
  "source": "meta"
}