|
I've read that a neural network with zero hidden layers is the same as a linear function. What function would that be though? Is the function known? Given a neural network with 0 hidden layers and 3 inputs, what linear function should I use to approximate that? Is there anyway to get that function or is it unknown? Thanks. |
|
Depending on the particulars, a single node or single layer of nodes would be more or less equivalent to a linear regression (unbounded real output), linear discriminant (class output) or a logistic regression (probability output). For a three-input linear model, each node would have the form: OUTPUT = W1 * INPUT1 + W2 * INPUT2 + W3 * INPUT3 ...where W1, W2 and W3 represent the model coefficients which could be fit the conventional way using least squares linear regression. The classification and probability models would be similar, with some type of transformation applied to the linear output, and the fitting process would obviously be different. @Will Thanks!
(Jan 30 '11 at 08:01)
CWhite
|
