Because of added more classifications, instead of matrix W, now we have matrix W1, W2, W3,W4. Does it mean instead of two dimensional matrix W, which is number of features * number of units in that layer, now the new matrix W1, W2, W3,W4 will be 3 dimensional, which is number of features * number of units in that layer * number of classes?
Here w_1, w_2, w_3, w_4 are vectors - the columns of matrix W. The number of columns corresponds to the number of classes in the output layer.
1 Like