{
  "id": 335844,
  "title": "Simple Silver Solution.",
  "content": "<p>Here is my solution for this competition. The code is <a href=\"https://www.kaggle.com/code/dongkyunkim/simple-silver-solution-foursquare-inference\" target=\"_blank\">here</a></p>\n<p>It received a silver medal without any ensembling or complicated feature engineering.</p>\n<p>The pipeline is is simple:</p>\n<ol>\n<li>Train xlmroberta with ArcFace Loss</li>\n<li>Use the cos sims from the xlmroberta + coordinate distance to extract match candidates</li>\n<li>Add features (cos sim, distance, lcs, tfidf, etc…)</li>\n<li>Train a lightgbm model (with flaml hyperparameter optimization) to select the correct candidates as a binary classification task.</li>\n<li>Do 2-3 on the test data and inference with lgbm</li>\n</ol>",
  "comments": [
    {
      "id": 1847832,
      "author_name": "DeepUnderstanding",
      "author_url": "",
      "post_date": "2022-07-08T07:24:24.250000",
      "content": "<p><code>Use the cos sims from the xlmroberta + coordinate distance to extract match candidates</code><br>\nCould you briefly explain this part</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1848828,
          "author_name": "Dongkyun Kim",
          "author_url": "",
          "post_date": "2022-07-08T23:59:24.317000",
          "content": "<p>Okay so I simply concat all string columns<br>\ninto one large sentence, infer word embeddings with the trained xlmroberta, then do k nearest neighbor retrieval with k=10 with cosine similarity. I also do k nearest neighbor retrieval with the longitude and latitude. The combined are my final candidates.</p>",
          "votes": 2,
          "replies": []
        }
      ]
    },
    {
      "id": 1851090,
      "author_name": "Dongkyun Kim",
      "author_url": "",
      "post_date": "2022-07-11T02:40:43.190000",
      "content": "<p>Training code is <a href=\"https://github.com/dongkyuk/Kaggle-Foursquare\" target=\"_blank\">here</a> if anyone is interested </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1848718,
      "author_name": "The Devastator",
      "author_url": "",
      "post_date": "2022-07-08T21:00:51.077000",
      "content": "<p>I just want to say that I am very impressed with your solution! Your pipeline is very simple and elegant.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 1858791,
      "author_name": "Youngjune Jeong",
      "author_url": "",
      "post_date": "2022-07-17T08:00:03.387000",
      "content": "<p>Thank you for sharing!</p>",
      "votes": 1,
      "replies": []
    }
  ],
  "comment_count": 5,
  "messages": [
    {
      "id": 1847686,
      "postDate": "2022-07-08T04:47:09.580Z",
      "content": "<p>Here is my solution for this competition. The code is <a href=\"https://www.kaggle.com/code/dongkyunkim/simple-silver-solution-foursquare-inference\" target=\"_blank\">here</a></p>\n<p>It received a silver medal without any ensembling or complicated feature engineering.</p>\n<p>The pipeline is is simple:</p>\n<ol>\n<li>Train xlmroberta with ArcFace Loss</li>\n<li>Use the cos sims from the xlmroberta + coordinate distance to extract match candidates</li>\n<li>Add features (cos sim, distance, lcs, tfidf, etc…)</li>\n<li>Train a lightgbm model (with flaml hyperparameter optimization) to select the correct candidates as a binary classification task.</li>\n<li>Do 2-3 on the test data and inference with lgbm</li>\n</ol>",
      "rawMarkdown": "Here is my solution for this competition. The code is [here](https://www.kaggle.com/code/dongkyunkim/simple-silver-solution-foursquare-inference)\n\nIt received a silver medal without any ensembling or complicated feature engineering.\n\nThe pipeline is is simple:\n\n1. Train xlmroberta with ArcFace Loss\n2. Use the cos sims from the xlmroberta + coordinate distance to extract match candidates\n3. Add features (cos sim, distance, lcs, tfidf, etc...)\n4. Train a lightgbm model (with flaml hyperparameter optimization) to select the correct candidates as a binary classification task.\n5. Do 2-3 on the test data and inference with lgbm",
      "votes": 10
    },
    {
      "id": 1847832,
      "postDate": "2022-07-08T07:24:24.250Z",
      "content": "<p><code>Use the cos sims from the xlmroberta + coordinate distance to extract match candidates</code><br>\nCould you briefly explain this part</p>",
      "rawMarkdown": "`Use the cos sims from the xlmroberta + coordinate distance to extract match candidates`\nCould you briefly explain this part",
      "votes": 2,
      "replies": [
        {
          "id": 1848828,
          "postDate": "2022-07-08T23:59:24.317Z",
          "content": "<p>Okay so I simply concat all string columns<br>\ninto one large sentence, infer word embeddings with the trained xlmroberta, then do k nearest neighbor retrieval with k=10 with cosine similarity. I also do k nearest neighbor retrieval with the longitude and latitude. The combined are my final candidates.</p>",
          "rawMarkdown": "Okay so I simply concat all string columns\ninto one large sentence, infer word embeddings with the trained xlmroberta, then do k nearest neighbor retrieval with k=10 with cosine similarity. I also do k nearest neighbor retrieval with the longitude and latitude. The combined are my final candidates.",
          "votes": 2
        }
      ]
    },
    {
      "id": 1851090,
      "postDate": "2022-07-11T02:40:43.190Z",
      "content": "<p>Training code is <a href=\"https://github.com/dongkyuk/Kaggle-Foursquare\" target=\"_blank\">here</a> if anyone is interested </p>",
      "rawMarkdown": "Training code is [here](https://github.com/dongkyuk/Kaggle-Foursquare) if anyone is interested ",
      "votes": 1
    },
    {
      "id": 1848718,
      "postDate": "2022-07-08T21:00:51.077Z",
      "content": "<p>I just want to say that I am very impressed with your solution! Your pipeline is very simple and elegant.</p>",
      "rawMarkdown": "I just want to say that I am very impressed with your solution! Your pipeline is very simple and elegant.\n",
      "votes": 1
    },
    {
      "id": 1858791,
      "postDate": "2022-07-17T08:00:03.387Z",
      "content": "<p>Thank you for sharing!</p>",
      "rawMarkdown": "Thank you for sharing!",
      "votes": 1
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-23T22:19:43Z",
  "competition_id": 35476,
  "competition_slug": "foursquare-location-matching",
  "team_id": 8596803,
  "solution_url": "/competitions/foursquare-location-matching/writeups/dongkyun-kim-simple-silver-solution"
}