Hi,
I have a question regarding the number of nodes/neurons for the output layer for classification CLP problem. To learn the representation, you can have the access to the whole dataset, and the meta objective is minimized by S_test sampled from \tau. Does it mean that the size of output layer is always fixed regardless how many classes the network has seemed so far?
I see the output size is always 1000 from the following code？
"config": {"out": 1000, "in": 9 * channels}
Thanks.
Hi,
I have a question regarding the number of nodes/neurons for the output layer for classification CLP problem. To learn the representation, you can have the access to the whole dataset, and the meta objective is minimized by S_test sampled from \tau. Does it mean that the size of output layer is always fixed regardless how many classes the network has seemed so far?
I see the output size is always 1000 from the following code？
"config": {"out": 1000, "in": 9 * channels}Thanks.