continue.functions: Continue functions

In xrobin/DeepLearning: Deep Learning of neural networks

Description

Functions to stop (by returning FALSE) the training if it has converged.

continue.function.exponential fits an exponential to the error and return TRUE if the function hasn't converged (or in case of doubt), FALSE if the function has converged or if the data couldn't be fitted (plateau reached or no exponential fit). As a side effect, this function plots the error and the fit, and prints a summary of the fit on the console

continue.function.always always returns true so that the training carries on until maxiters is reached

Usage

continue.function.exponential(error, iter, batchsize, maxiters,
  layer = 0, ic = AIC)

continue.function.always(error, iter, batchsize, maxiters, layer = 0)

continue.function.exponential.aic(error, iter, batchsize, maxiters,
  layer = 0)

continue.function.exponential.bic(error, iter, batchsize, maxiters,
  layer = 0)

continue.function.random(error, iter, batchsize, maxiters, layer = 0)

Arguments

error	a vector of the errors along the training
iter, batchsize	current iteration number and batchsize.
maxiters	maximum number of iterations
layer	during RBM pre-training, which layer is being pre-trained. Otherwise, 0.
ic	the Information Criterion to use, typically AIC or BIC

Value

boolean (see description)

xrobin/DeepLearning documentation built on Sept. 18, 2020, 5:23 a.m.