{
  "id": 448839,
  "title": "Predict at celltype-drug level or celltype-drug-gene level?",
  "url": "/competitions/open-problems-single-cell-perturbations/discussion/448839",
  "author_name": "Edward",
  "post_date": "2023-10-21T18:27:30.121000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I have a question about in which level you are training your model? I take each gene as a sample, and input the celltype token, drug token, and gene token to predict the output, but my best LB is just 0.616. And <a href=\"https://www.kaggle.com/Kishan\" target=\"_blank\">@Kishan</a> Vavdara used all genes of a cell type as a sample and took 614 samples to reach a 0.604 LB. I don't know how you trained your model and I am very confused.😢</p>",
  "messages": [
    {
      "id": 2491579,
      "postDate": "2023-10-21T18:27:30.120Z",
      "content": "<p>I have a question about in which level you are training your model? I take each gene as a sample, and input the celltype token, drug token, and gene token to predict the output, but my best LB is just 0.616. And <a href=\"https://www.kaggle.com/Kishan\" target=\"_blank\">@Kishan</a> Vavdara used all genes of a cell type as a sample and took 614 samples to reach a 0.604 LB. I don't know how you trained your model and I am very confused.😢</p>",
      "rawMarkdown": "I have a question about in which level you are training your model? I take each gene as a sample, and input the celltype token, drug token, and gene token to predict the output, but my best LB is just 0.616. And @Kishan Vavdara used all genes of a cell type as a sample and took 614 samples to reach a 0.604 LB. I don't know how you trained your model and I am very confused.😢",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2491579": "I have a question about in which level you are training your model? I take each gene as a sample, and input the celltype token, drug token, and gene token to predict the output, but my best LB is just 0.616. And @Kishan Vavdara used all genes of a cell type as a sample and took 614 samples to reach a 0.604 LB. I don't know how you trained your model and I am very confused.😢"
  }
}