{
  "id": 262446,
  "title": "Augmentation and LR_scheduler have any relations?",
  "url": "/competitions/seti-breakthrough-listen/discussion/262446",
  "author_name": "Vatsal Mavani",
  "post_date": "2021-08-06T17:37:43.504000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I have noticed that some augmentations work well with below lr_scheduler and Ranger optimizer while some give the worst performance. Anyone knows about such relation or how to do analysis for this. OR it just changes on datasets and we have to do experiments for it.</p>\n<pre><code>def lrfn(epoch):\n        if (epoch &lt; lr_ramp_ep):\n            lr = (lr_max - lr_start) / lr_ramp_ep * epoch + lr_start\n        elif (epoch &lt; lr_ramp_ep + lr_sus_ep):\n            lr = lr_max\n        else:\n            lr = (lr_max - lr_min) * lr_decay**(epoch - lr_ramp_ep - lr_sus_ep) + lr_min\n        return lr\n</code></pre>",
  "messages": [
    {
      "id": 1455813,
      "postDate": "2021-08-06T17:37:43.503Z",
      "content": "<p>I have noticed that some augmentations work well with below lr_scheduler and Ranger optimizer while some give the worst performance. Anyone knows about such relation or how to do analysis for this. OR it just changes on datasets and we have to do experiments for it.</p>\n<pre><code>def lrfn(epoch):\n        if (epoch &lt; lr_ramp_ep):\n            lr = (lr_max - lr_start) / lr_ramp_ep * epoch + lr_start\n        elif (epoch &lt; lr_ramp_ep + lr_sus_ep):\n            lr = lr_max\n        else:\n            lr = (lr_max - lr_min) * lr_decay**(epoch - lr_ramp_ep - lr_sus_ep) + lr_min\n        return lr\n</code></pre>",
      "rawMarkdown": "I have noticed that some augmentations work well with below lr_scheduler and Ranger optimizer while some give the worst performance. Anyone knows about such relation or how to do analysis for this. OR it just changes on datasets and we have to do experiments for it.\n\n```\ndef lrfn(epoch):\n        if (epoch < lr_ramp_ep):\n            lr = (lr_max - lr_start) / lr_ramp_ep * epoch + lr_start\n        elif (epoch < lr_ramp_ep + lr_sus_ep):\n            lr = lr_max\n        else:\n            lr = (lr_max - lr_min) * lr_decay**(epoch - lr_ramp_ep - lr_sus_ep) + lr_min\n        return lr\n\n```"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1455813": "I have noticed that some augmentations work well with below lr_scheduler and Ranger optimizer while some give the worst performance. Anyone knows about such relation or how to do analysis for this. OR it just changes on datasets and we have to do experiments for it.\n\n```\ndef lrfn(epoch):\n        if (epoch < lr_ramp_ep):\n            lr = (lr_max - lr_start) / lr_ramp_ep * epoch + lr_start\n        elif (epoch < lr_ramp_ep + lr_sus_ep):\n            lr = lr_max\n        else:\n            lr = (lr_max - lr_min) * lr_decay**(epoch - lr_ramp_ep - lr_sus_ep) + lr_min\n        return lr\n\n```"
  }
}