{
  "id": 475593,
  "title": "Best models for HMS comp? ",
  "url": "/competitions/hms-harmful-brain-activity-classification/discussion/475593",
  "author_name": "",
  "post_date": "2024-02-09T05:20:12.646084100Z",
  "votes": 6,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I have seen a number of different strategies work fairly well in this competition so far and I was curious what sort of models people are finding success with? So far I am seeing EfficientNetB0, EfficientNetB2, ResNet34d, and ResNet1D models. I would expect a number of other Timm models to work as well and haven’t gotten around to try them too much as the Sennet comp just ended. </p>",
  "messages": [
    {
      "id": "2643780",
      "postDate": "02/09/2024 05:20:12",
      "content": "<p>I have seen a number of different strategies work fairly well in this competition so far and I was curious what sort of models people are finding success with? So far I am seeing EfficientNetB0, EfficientNetB2, ResNet34d, and ResNet1D models. I would expect a number of other Timm models to work as well and haven’t gotten around to try them too much as the Sennet comp just ended. </p>",
      "rawMarkdown": "I have seen a number of different strategies work fairly well in this competition so far and I was curious what sort of models people are finding success with? So far I am seeing EfficientNetB0, EfficientNetB2, ResNet34d, and ResNet1D models. I would expect a number of other Timm models to work as well and haven’t gotten around to try them too much as the Sennet comp just ended.",
      "votes": null
    },
    {
      "id": "2644244",
      "postDate": "02/09/2024 10:47:48",
      "content": "<p>EfficientNet Series: EfficientNet is a network designed to be efficient in how it scales the size and performance of its models. I thought this was because there are models of various sizes, from small models to large models, so it is easy to apply them to various datasets.<br>\nI'm translating from Korean to English, so I hope you understand even if the sentence structure is awkward.</p>",
      "rawMarkdown": "EfficientNet Series: EfficientNet is a network designed to be efficient in how it scales the size and performance of its models. I thought this was because there are models of various sizes, from small models to large models, so it is easy to apply them to various datasets.\nI'm translating from Korean to English, so I hope you understand even if the sentence structure is awkward.",
      "votes": null
    },
    {
      "id": "2644591",
      "postDate": "02/09/2024 15:28:35",
      "content": "<p>I think you'd rather spend time in data pre-processing, building a better training/inference approach, .. rather than simply changing the encoder's architecture, because it only gets you so far. Just my 2 cents.</p>",
      "rawMarkdown": "I think you'd rather spend time in data pre-processing, building a better training/inference approach, .. rather than simply changing the encoder's architecture, because it only gets you so far. Just my 2 cents.",
      "votes": null
    },
    {
      "id": "2644700",
      "postDate": "02/09/2024 16:20:01",
      "content": "<p>I would agree, seems like there is a lot of signal in the data from my research so far! </p>",
      "rawMarkdown": "I would agree, seems like there is a lot of signal in the data from my research so far!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2644244,
      "author_name": "tsebymyrt",
      "author_url": "",
      "post_date": "02/09/2024 10:47:48",
      "content": "<p>EfficientNet Series: EfficientNet is a network designed to be efficient in how it scales the size and performance of its models. I thought this was because there are models of various sizes, from small models to large models, so it is easy to apply them to various datasets.<br>\nI'm translating from Korean to English, so I hope you understand even if the sentence structure is awkward.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2644591,
      "author_name": "fahmiayari",
      "author_url": "",
      "post_date": "02/09/2024 15:28:35",
      "content": "<p>I think you'd rather spend time in data pre-processing, building a better training/inference approach, .. rather than simply changing the encoder's architecture, because it only gets you so far. Just my 2 cents.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2644700,
          "author_name": "cody11null",
          "author_url": "",
          "post_date": "02/09/2024 16:20:01",
          "content": "<p>I would agree, seems like there is a lot of signal in the data from my research so far! </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2643780": "I have seen a number of different strategies work fairly well in this competition so far and I was curious what sort of models people are finding success with? So far I am seeing EfficientNetB0, EfficientNetB2, ResNet34d, and ResNet1D models. I would expect a number of other Timm models to work as well and haven’t gotten around to try them too much as the Sennet comp just ended.",
    "2644244": "EfficientNet Series: EfficientNet is a network designed to be efficient in how it scales the size and performance of its models. I thought this was because there are models of various sizes, from small models to large models, so it is easy to apply them to various datasets.\nI'm translating from Korean to English, so I hope you understand even if the sentence structure is awkward.",
    "2644591": "I think you'd rather spend time in data pre-processing, building a better training/inference approach, .. rather than simply changing the encoder's architecture, because it only gets you so far. Just my 2 cents.",
    "2644700": "I would agree, seems like there is a lot of signal in the data from my research so far!"
  },
  "source": "meta"
}