728x90
1. Why do we need activation function?
learns the function f which best determines the relationship between input x and output y.
-> Unfortunately this relation in most cases is non-linear
For example a single sigmoid function looks like this.
Now, by joining multiple such sigmoid curves at different layers of FNN,
you can almost figure out any complex relations between input and output.
- 이론적으로는 2층으로 network를 쌓더라도 임의의 연속함수를 근사할 수 있다.
- 하지만 층을 많이 쌓을 수록 근사에 필요한 neuron의 수가 훨신 많이 줄어 들기 때문에 효율적 학습이 가능하다.
728x90
반응형
'boostcamp AI tech > boostcamp AI' 카테고리의 다른 글
| BPTT for RNN (0) | 2023.08.26 |
|---|---|
| Recurrent Neural Networks (0) | 2023.08.23 |
| Convolutional Neural Network (0) | 2023.08.23 |
| Autoencoder (0) | 2023.08.17 |
| Cost functions and Gradient descent (0) | 2023.08.17 |
