Learn abou from the vide lecture:
Why to learn these operations first? In fact, PyTorch is not only a library enabeling you to perform "fit a model". Before traning a model, let's make a step back !
On granular level, deep learning is s rigorious application of linear algebra. Remember, even your training data will need to be respresented as a tensor.
Before training a model, you will need to transform your data into a format and shape your neural network can digest. Look at this official PyTorch documentation and l**earn how easy PyTorch makes it for you to perform these operations.
Dot Product and Matrix Multiplication
Matrix Transpose Look here to have immediate idea about what this opperation looks like. In fact, matrix transposition (after matrix multiplication) is the most essential operation for making a NN architecture work. The main reason is that it makes it possible to multiply two matrices like: and . This opperation happens immediatelly when input training data (represented as a tensor) flows to fully connected layer.