Closed ALEEF02 closed 4 years ago
sounds like you're overfitting your dataset. extrapolate new point may result different results, due the stochastic nature of neural networks.
I fixed the issue a couple weeks back. The issue was that the y
in trainer.train(x, y);
is not supposed to be a covnetjs.Vol
. It should just be put in as an array of output values.
I have the following code to make a net with 4 inputs, one hidden layer, and 4 outputs. When I try to pull a prediction from the data after training, the values are always wildly different on every reload. Am I doing something wrong?