Package mklab.JGNN.nn

Interface Optimizer

All Known Implementing Classes:
Adam, BatchOptimizer, GradientDescent, Regularization
public interface Optimizer
Provides an interface for training tensors. Has a reset() method that starts potential training memory from scratch. Has an update(Tensor, Tensor) method that, given a current Tensor and a gradient operates on the former and adjusts its value.
Author:
Emmanouil Krasanakis

  • Method Summary

    Modifier and Type
    Method
    Description
    default void
    reset()
    Resets (and lets the garbage collector free) optimizer memory.
    void
    update(Tensor value, Tensor gradient)
    In-place updates the value of a tensor given its gradient.

  • Method Details

    • update

      void update(Tensor value, Tensor gradient)
      In-place updates the value of a tensor given its gradient. Some optimizers (e.g. Adama) require the exact same tensor instance to be provided so as to keep track of its optimization progress. The library makes sure to keep this constraint.
      Parameters:
      value - The tensor to update.
      gradient - The tensor's gradient.

    • reset

      default void reset()
      Resets (and lets the garbage collector free) optimizer memory. Should be called at the beginning of training (not after each epoch).