Let's say the second to last layer of my neural network outputs a tensor with shape [7].
How would I make my output layer be only a subset of the previous 7 outputs, for example an array of the values at index 0, 3, and 6?
For example, if my second to last output is [0.2, 1, 6, 1, 2, 0.87, 1.7], how could I get the last layer to output [0.2, 1, 1.7]?
