{
  "id": 322494,
  "title": "\"recall method\" terminology",
  "url": "/competitions/h-and-m-personalized-fashion-recommendations/discussion/322494",
  "author_name": "",
  "post_date": "2022-05-02T14:38:16.276640600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I see a few discussion posts talk about \"recall methods\".</p>\n<p>As far as I can tell, what this means is a strategy to generate candidate items for each customer for a given week: some of these items the customer will have bought (positive samples, label will be 1), whilst others the customer won't have bought (negative samples, label will be 0).</p>\n<p>Where does this terminology come from? I've looked at the lgbmranker docs, and searched around, by \"recall methods\" doesn't return anything relevant. The \"learning to rank\" Wiki page doesn't mention recall.</p>",
  "messages": [
    {
      "id": "1774881",
      "postDate": "05/02/2022 14:38:16",
      "content": "<p>I see a few discussion posts talk about \"recall methods\".</p>\n<p>As far as I can tell, what this means is a strategy to generate candidate items for each customer for a given week: some of these items the customer will have bought (positive samples, label will be 1), whilst others the customer won't have bought (negative samples, label will be 0).</p>\n<p>Where does this terminology come from? I've looked at the lgbmranker docs, and searched around, by \"recall methods\" doesn't return anything relevant. The \"learning to rank\" Wiki page doesn't mention recall.</p>",
      "rawMarkdown": "I see a few discussion posts talk about \"recall methods\".\n\nAs far as I can tell, what this means is a strategy to generate candidate items for each customer for a given week: some of these items the customer will have bought (positive samples, label will be 1), whilst others the customer won't have bought (negative samples, label will be 0).\n\nWhere does this terminology come from? I've looked at the lgbmranker docs, and searched around, by \"recall methods\" doesn't return anything relevant. The \"learning to rank\" Wiki page doesn't mention recall.",
      "votes": null
    },
    {
      "id": "1775110",
      "postDate": "05/02/2022 17:51:22",
      "content": "<p>I think \"recall methods\" are more commonly referred to as \"generating candidates\" or \"candidate generation\" for the recsys field, but I might be wrong</p>",
      "rawMarkdown": "I think \"recall methods\" are more commonly referred to as \"generating candidates\" or \"candidate generation\" for the recsys field, but I might be wrong",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1775110,
      "author_name": "kennyxie",
      "author_url": "",
      "post_date": "05/02/2022 17:51:22",
      "content": "<p>I think \"recall methods\" are more commonly referred to as \"generating candidates\" or \"candidate generation\" for the recsys field, but I might be wrong</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1774881": "I see a few discussion posts talk about \"recall methods\".\n\nAs far as I can tell, what this means is a strategy to generate candidate items for each customer for a given week: some of these items the customer will have bought (positive samples, label will be 1), whilst others the customer won't have bought (negative samples, label will be 0).\n\nWhere does this terminology come from? I've looked at the lgbmranker docs, and searched around, by \"recall methods\" doesn't return anything relevant. The \"learning to rank\" Wiki page doesn't mention recall.",
    "1775110": "I think \"recall methods\" are more commonly referred to as \"generating candidates\" or \"candidate generation\" for the recsys field, but I might be wrong"
  },
  "source": "meta"
}