NN - neurons with "interval rules"
No.11889666 ViewReplyOriginalReport
Quoted By: >>11889675 >>11889679 >>11890056 >>11890085
Suppose you have a bunch of sequences corresponding to a bunch of numbers.
inputs = [[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
...
[1, 1, 1, 1, 1, 0, 1, 1, 0]]
outputs = [-0.11, 0.72, -0.83, 1.13, ..., 0.45]
I am using a single neuron, nothing special, to train it on the data above, but there is a problem. A newly introduced sequence: S = [1, 1, 1, 1, 1, 0, 0, 0, 0], can only exists within a given interval, 0.0 < S < 0.1, and my neuron that i trained on the data above does of course not know this. Without this "rule" in mind the trained neuron thinks that sequence: S = 0.35. I need to somehow let it know that sequence: S, can only take on values within a given interval, question is; how? If at all possible... Any ideas on possible solutions?
inputs = [[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
[1, 1, 1, 1, 1, 0, 1, 1, 0],
...
[1, 1, 1, 1, 1, 0, 1, 1, 0]]
outputs = [-0.11, 0.72, -0.83, 1.13, ..., 0.45]
I am using a single neuron, nothing special, to train it on the data above, but there is a problem. A newly introduced sequence: S = [1, 1, 1, 1, 1, 0, 0, 0, 0], can only exists within a given interval, 0.0 < S < 0.1, and my neuron that i trained on the data above does of course not know this. Without this "rule" in mind the trained neuron thinks that sequence: S = 0.35. I need to somehow let it know that sequence: S, can only take on values within a given interval, question is; how? If at all possible... Any ideas on possible solutions?
