Binary prediction is cool and fun to play with. One can implement a NN to manage business logic, classify objects in images or simply help finding a decision.
Another use case for NNs are numeric predictions, which we want to add to the library.
To achieve that, one needs to
add normalization to the datasets, easy!
normalizedX = (x-min(x))/(max(x)-min(x))
replace sigmoid with identity for activation
x = x
and obviously it's derivative
x' = 1
Implement Predict(input) method for NN
at the end implement a test, use sin() or cos() to generate datasets and for testing
Binary prediction is cool and fun to play with. One can implement a NN to manage business logic, classify objects in images or simply help finding a decision. Another use case for NNs are numeric predictions, which we want to add to the library. To achieve that, one needs to
and obviously it's derivative
Predict(input)
method for NNat the end implement a test, use
sin()
orcos()
to generate datasets and for testing