반응형
    
    
    
  
                              Notice
                              
                          
                        
                          
                          
                            Recent Posts
                            
                        
                          
                          
                            Recent Comments
                            
                        
                          
                          
                            Link
                            
                        
                    | 일 | 월 | 화 | 수 | 목 | 금 | 토 | 
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | |||
| 5 | 6 | 7 | 8 | 9 | 10 | 11 | 
| 12 | 13 | 14 | 15 | 16 | 17 | 18 | 
| 19 | 20 | 21 | 22 | 23 | 24 | 25 | 
| 26 | 27 | 28 | 29 | 30 | 31 | 
                            Tags
                            
                        
                          
                          - 회로
- 풀다운저항
- Backpropagation
- impl
- palindrome
- Features
- assignment1
- error
- softmax backpropagation
- 풀업저항
- pyTorch
- two-layer neural net
- pulldown
- neural net
- Circuit
- autoencoder
- Softmax
- assignment
- cs231n
- Solution
- KNN
- TensorFlow
- NotFoundError
- atmega128
- pullup
- backward pass
- leetcode
- Floating
- #9
- Big size image
                            Archives
                            
                        
                          
                          - Today
- Total
목록two-layer neural net (1)
코딩공부
 cs231n assignment1 - two-layer neural net
      
      
        cs231n assignment1 - two-layer neural net
        과제의 전체 코드는 https://github.com/Lee-daeho/cs231n/tree/master/assignment1에 업로드 되어 있다. 우선 softmax의 backpropagation을 설명해보고자 한다. 코드로는 다음과 같이 표현할 수 있다. 여기서 gradient를 계산하기 위해 backward pass(backpropagation)을 진행해야 하는데 이것을 자세히 풀어보았다. 이런 과정을 통해 각 input의 gradient를 구할 수 있다. 대부분의 코드는 기존의 과제 softmax나 linear classifier를 참조하면서 진행할 수 있으므로 마지막으로 Inline Question만 보겠다. 질문은 training accuracy와 testing accuracy의 차이가 크면 ..
        ML , DL (2019)/cs231n
        
        2020. 1. 10. 18:35