public Adam(learningRate as Double = 0.001, beta1 as Double = 0.9, beta2 as Double = 0.999, epsilon as Double = 1e-7)
Configures Adam without weight decay or AMSGrad.
| Parameter | Description |
|---|---|
learningRate | Positive finite step size. |
beta1 | First-moment decay in [0, 1). |
beta2 | Second-moment decay in [0, 1). |
epsilon | Positive finite stabilizer, outside the square root. |
optimizer = Adam<Float>(0.001, epsilon=1e-7)