I am training my Xception model with imagenet weights on Keras right now, on floydhub, and I am noticing that I am getting pretty good training accuracy(around 85 percent at about 10 epochs) but my validation accuracy is very low(about 20 percent). What do I do about this?
", "messages": [ { "id": "281689", "postDate": "02/12/2018 23:11:55", "content": "I am training my Xception model with imagenet weights on Keras right now, on floydhub, and I am noticing that I am getting pretty good training accuracy(around 85 percent at about 10 epochs) but my validation accuracy is very low(about 20 percent). What do I do about this?
", "rawMarkdown": "I am training my Xception model with imagenet weights on Keras right now, on floydhub, and I am noticing that I am getting pretty good training accuracy(around 85 percent at about 10 epochs) but my validation accuracy is very low(about 20 percent). What do I do about this?", "votes": null }, { "id": "282094", "postDate": "02/13/2018 16:12:32", "content": "This is a problem of high variance. Start by doing some regularization on your network. L2 regularization or dropout. That might improve validation accuracy.
", "rawMarkdown": "This is a problem of high variance. Start by doing some regularization on your network. L2 regularization or dropout. That might improve validation accuracy.", "votes": null }, { "id": "287191", "postDate": "02/23/2018 20:24:48", "content": "It means there is over-fitting of your model. Use augmentation to increase your data size, try different regularization techniques such as dropout, and slow down your learning rate.
", "rawMarkdown": "It means there is over-fitting of your model. Use augmentation to increase your data size, try different regularization techniques such as dropout, and slow down your learning rate.", "votes": null }, { "id": "1281033", "postDate": "04/22/2021 15:26:13", "content": "can anyone tell me why it shows 0 score for everyone in leaderboard @pukkinming @helper
", "rawMarkdown": "can anyone tell me why it shows 0 score for everyone in leaderboard @pukkinming @helper", "votes": null } ], "comments": [ { "id": 1281033, "author_name": "siddheshbalsaraf", "author_url": "", "post_date": "04/22/2021 15:26:13", "content": "can anyone tell me why it shows 0 score for everyone in leaderboard @pukkinming @helper
", "votes": null, "replies": [] }, { "id": 282094, "author_name": "helper", "author_url": "", "post_date": "02/13/2018 16:12:32", "content": "This is a problem of high variance. Start by doing some regularization on your network. L2 regularization or dropout. That might improve validation accuracy.
", "votes": null, "replies": [] }, { "id": 287191, "author_name": "pukkinming", "author_url": "", "post_date": "02/23/2018 20:24:48", "content": "It means there is over-fitting of your model. Use augmentation to increase your data size, try different regularization techniques such as dropout, and slow down your learning rate.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "281689": "I am training my Xception model with imagenet weights on Keras right now, on floydhub, and I am noticing that I am getting pretty good training accuracy(around 85 percent at about 10 epochs) but my validation accuracy is very low(about 20 percent). What do I do about this?", "282094": "This is a problem of high variance. Start by doing some regularization on your network. L2 regularization or dropout. That might improve validation accuracy.", "287191": "It means there is over-fitting of your model. Use augmentation to increase your data size, try different regularization techniques such as dropout, and slow down your learning rate.", "1281033": "can anyone tell me why it shows 0 score for everyone in leaderboard @pukkinming @helper" }, "source": "meta" }