keepdim
<aside> <img src="data:image/svg+xml;charset=utf-8;base64,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" alt="data:image/svg+xml;charset=utf-8;base64,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" width="40px" /> 차원을 유지할지 여부를 지정한다. kepdim=True이면, 결과 텐서가 원래 차원 구조를 유지하며, 유지하는 차원은 크기가 1로 설정된다. 반대로 keepdim=False이면, 결과 텐서에서 지정된 차원이 축소되어, 하나의 차원이 제거된다.
</aside>
t1 = torch.tensor([[10, 2, 3, 40], [5, 6, 7, 8], [9, 15, 25, 1]])col_argmax = torch.argmax(t1, dim=0) print(col_argmax) # tensor([0, 2, 2, 0])col_argmax_keepdim = torch.argmax(t1, dim=0, keepdim=True) print(col_argmax_keepdim) # tensor([[0, 2, 2, 0]])
[선형회귀방적식 중] X에 절편(bias) 항 추가
<aside> <img src="data:image/svg+xml;charset=utf-8;base64,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" alt="data:image/svg+xml;charset=utf-8;base64,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" width="40px" /> 이 작업을 하는 이유는 선형 회귀 모델에서 절편 w0를 포함한 가중치 벡터를 쉽게 계산하기 위해서이다. 만약 이 작업을 하지 않으면, 절편 w0를 계산하기 위해 추가적인 코드를 작성해야 한다. 하지만 이 방법을 사용하면, w0를 포함한 모든 가중치가 하나의 벡터로 표현될 수 있기 때문에 계산이 단순해진다.
</aside>