I'm not sure I understand something from this article:
In the context of linear classification of images, it is written:
Interpretation of linear classifiers as template matching: Another interpretation for the weights W is that each row of W corresponds to a template (or sometimes also called a prototype) for one of the classes. The score of each class for an image is then obtained by comparing each template with the image using an inner product (or dot product) one by one to find the one that “fits” best. With this terminology, the linear classifier is doing template matching, where the templates are learned.
But how could we use this as template matching for images with pixel values of 0? Which might happen also for mean centered data, as the image will be grey.
Also, not just for 0 pixel values, but low pixel values in general, will the weights be very large to compensate for low pixel values?(and then W as an image will not be similar by color to the images we learned from)