I have an issue that i couldn't fixed. I put my submission , it return a score less than 1% . However, my score estimation is about 96% based on data validation in my notebook. Besides , i make comparaison between my results and others in public notebook highly scored , i find nearly the same results. Please can someone give me a help how to fix this problem.\nThank you !
", "messages": [ { "id": "751545", "postDate": "02/20/2020 09:10:40", "content": "I have an issue that i couldn't fixed. I put my submission , it return a score less than 1% . However, my score estimation is about 96% based on data validation in my notebook. Besides , i make comparaison between my results and others in public notebook highly scored , i find nearly the same results. Please can someone give me a help how to fix this problem.\nThank you !
", "rawMarkdown": "I have an issue that i couldn't fixed. I put my submission , it return a score less than 1% . However, my score estimation is about 96% based on data validation in my notebook. Besides , i make comparaison between my results and others in public notebook highly scored , i find nearly the same results. Please can someone give me a help how to fix this problem.\nThank you !", "votes": null }, { "id": "751563", "postDate": "02/20/2020 09:29:57", "content": "I had a similiar problem and scored bottom 2% on my first attempt. It is likely that something with your data preprocessing is wrong in the submission (in my case, I forgot to invert the images as I had done in training). Your training data might also be incorrectly aligned so that it performs decent on the incorrectly aligned training data and obviously cannot transfer that over to the test data. In any event, you should probably compare how you preprocessed the data in training with how you preprocessed it in your submission.
", "rawMarkdown": "I had a similiar problem and scored bottom 2% on my first attempt. It is likely that something with your data preprocessing is wrong in the submission (in my case, I forgot to invert the images as I had done in training). Your training data might also be incorrectly aligned so that it performs decent on the incorrectly aligned training data and obviously cannot transfer that over to the test data. In any event, you should probably compare how you preprocessed the data in training with how you preprocessed it in your submission.", "votes": null } ], "comments": [ { "id": 751563, "author_name": "larswigger", "author_url": "", "post_date": "02/20/2020 09:29:57", "content": "I had a similiar problem and scored bottom 2% on my first attempt. It is likely that something with your data preprocessing is wrong in the submission (in my case, I forgot to invert the images as I had done in training). Your training data might also be incorrectly aligned so that it performs decent on the incorrectly aligned training data and obviously cannot transfer that over to the test data. In any event, you should probably compare how you preprocessed the data in training with how you preprocessed it in your submission.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "751545": "I have an issue that i couldn't fixed. I put my submission , it return a score less than 1% . However, my score estimation is about 96% based on data validation in my notebook. Besides , i make comparaison between my results and others in public notebook highly scored , i find nearly the same results. Please can someone give me a help how to fix this problem.\nThank you !", "751563": "I had a similiar problem and scored bottom 2% on my first attempt. It is likely that something with your data preprocessing is wrong in the submission (in my case, I forgot to invert the images as I had done in training). Your training data might also be incorrectly aligned so that it performs decent on the incorrectly aligned training data and obviously cannot transfer that over to the test data. In any event, you should probably compare how you preprocessed the data in training with how you preprocessed it in your submission." }, "source": "meta" }