Should AI training datasets include controversial or sensitive topics, or is it too risky?
학생이 쓴 원문
It doesn\'t essential a dataset includes sensitive samples. But it\'s common in deep learning field. And these sensitive data can occurs problem that deep learning network overfit to these samples. If network is trained focusing with these sensitive data, it can\'t work well with normal data. So we called these data as \'outlier\'. For using these outlier with normal data, we need to use pre-processing step before put into network. Pre-processing includes normalization, denoising, and scaling. It can help reducing remarkable characteristics of the outliers.
강사 첨삭
You demonstrated strong technical understanding of deep learning concepts such as overfitting, outliers, and pre-processing. Your reasoning was logical and well connected to the question about sensitive data. To improve further, focus on grammar structure, article usage, and natural academic phrasing. With small adjustments, this answer sounds clear, professional, and well-informed. Excellent work!
-Teacher Wela
But it\'s common in deep learning field.
>> However, it is common in the deep learning field.
And these sensitive data can occurs problem that deep learning network overfit to these samples.
>> These sensitive data can cause problems in which a deep learning network overfits to these samples.
If network is trained focusing with these sensitive data, it can\'t work well with normal data.
>> If a network is trained while focusing on these sensitive data, it may not work well with normal data.
So we called these data as \'outlier\'.
>> Therefore, we call these data outliers.
For using these outlier with normal data, we need to use pre-processing step before put into network.
>> To use these outliers with normal data, we need to apply a pre-processing step before putting them into the network.
It can help reducing remarkable characteristics of the outliers.
>> These techniques can help reduce the extreme characteristics of the outliers.
파워잉글리쉬 영어첨삭교정 서비스에 실제로 접수된 글입니다. 작성자와 글에 등장하는 사람·회사 이름, 연락처 등 개인을 알아볼 수 있는 정보는 모두 지웠습니다. 문법 오류는 학습 자료로서의 가치를 위해 원문 그대로 두었습니다.