![](https://img1.daumcdn.net/thumb/R750x0/?scode=mtistory2&fname=https%3A%2F%2Fblog.kakaocdn.net%2Fdn%2FdJg4iv%2FbtrOCZk8qVA%2FeNhFCqLkLxdpsoBJqS0tK1%2Fimg.png)
1. Why Deep Representations? Intuition about deep representation CNN에서 이미지 분석 과정은 위와 같다. 작은 것들을(edge) 먼저 분석하고 이것으로부터 점점 큰 조각들을 합쳐 원래의 모습을 구성한다. speech recognition에도 적용 가능하다 low level의 speech sound phonemes words sentence, phrase Circuit theory and deep learning Informally: There are functions you can compute with a "small" L-layer deep neural network that shallower networks require exponentially..