The learning framework sends over individual instances for application.
The main reason for this is that this is necessary if we want to use the special feature that gives us the classification of the previous instance.
We could restrict the batch size to 1 only if we actually use that feature, but for now we stick with batch size 1.
In that case, using the GPU is likely to be slower because of the additional transfer overhead, so for now, lets always use the CPU.
The learning framework sends over individual instances for application. The main reason for this is that this is necessary if we want to use the special feature that gives us the classification of the previous instance.
We could restrict the batch size to 1 only if we actually use that feature, but for now we stick with batch size 1.
In that case, using the GPU is likely to be slower because of the additional transfer overhead, so for now, lets always use the CPU.