{
  "id": 608319,
  "title": "Possible GNN Approach",
  "url": "/competitions/MABe-mouse-behavior-detection/discussion/608319",
  "author_name": "Tom",
  "post_date": "2025-09-19T04:25:56.597000",
  "votes": 21,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Social behavior structures might be well represented as graphs, since they also provide coordinates that allow us to utilize positional information. Here are some initial ideas for possible graph structures:</p>\n<ol>\n<li><p>Naive homogeneous graph – representing mice, with all nodes connected across all frames.</p></li>\n<li><p>Heterogeneous graph – representing mice and their body parts, incorporating positional and tracking information.</p></li>\n<li><p>Two-stage approach – first classify clear social behaviors, then apply label propagation on the graph to observe what label structures emerge.</p></li>\n</ol>",
  "messages": [
    {
      "id": 3291140,
      "postDate": "2025-09-19T04:25:56.597Z",
      "content": "<p>Social behavior structures might be well represented as graphs, since they also provide coordinates that allow us to utilize positional information. Here are some initial ideas for possible graph structures:</p>\n<ol>\n<li><p>Naive homogeneous graph – representing mice, with all nodes connected across all frames.</p></li>\n<li><p>Heterogeneous graph – representing mice and their body parts, incorporating positional and tracking information.</p></li>\n<li><p>Two-stage approach – first classify clear social behaviors, then apply label propagation on the graph to observe what label structures emerge.</p></li>\n</ol>",
      "rawMarkdown": "Social behavior structures might be well represented as graphs, since they also provide coordinates that allow us to utilize positional information. Here are some initial ideas for possible graph structures:\n\n1. Naive homogeneous graph – representing mice, with all nodes connected across all frames.\n\n\n2. Heterogeneous graph – representing mice and their body parts, incorporating positional and tracking information.\n\n\n3. Two-stage approach – first classify clear social behaviors, then apply label propagation on the graph to observe what label structures emerge.",
      "votes": 21
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3291140": "Social behavior structures might be well represented as graphs, since they also provide coordinates that allow us to utilize positional information. Here are some initial ideas for possible graph structures:\n\n1. Naive homogeneous graph – representing mice, with all nodes connected across all frames.\n\n\n2. Heterogeneous graph – representing mice and their body parts, incorporating positional and tracking information.\n\n\n3. Two-stage approach – first classify clear social behaviors, then apply label propagation on the graph to observe what label structures emerge."
  }
}