Shapes Dataset

• Convention is that the first dimension of the dataset belongs to the features and the second to the number of samples. The operations can then be systematically applied from the left.
• Exception!, in TensorFlow and PyTorch it is the other way around!
• $(\#features,\#samples)$