Changing the dimension/shape of a tensor

[jiachenpauljiang]

I have coded deep learning models in the past, but I never fully understood them. That's why I am learning it again from the scratch. I am about halfway down the first notebook, and this might be a really fundamental math question. When we change/reshape a tensor, say in order to do a matrix multiplication between two tensors with the same shape, the value of the tensor does not change although the dimension/shape has been changed. Is this correct? For example, the value of [[1,2,3], [1,2,3]] is the same as [[1,1], [2,2], [3,3]]? Thanks a lot!

[sriram-gupta]

what do you mean when you say " the value doesn't change but the dimension changed ". If you are talking in terms of what it represent it totally changes when we do any linear operation on it , be it transpose also.

[mrdbourke]

Yes @jiachenpauljiang , you're correct.

You can reshape/tensor a tensor without changing the underlying values.

You are only rearranging the values or the axes/dimensions.

Rearranging tensors can help with various manipulations or calculations (e.g. matrix multiplication).