sample_lr_values_log_scale: Randomly sample values for learning rate parameter 'alpha' on...

In ifrit98/R2deepR:

Description

Computes minimum values on a log scale as 'a <- log(lo, 10)', and 'b <- log(hi, 10)'. Calls to 'runif()' are scaled to the appropriate range by 'a' and are checked to be out of bounds against '-b'. Values of 'alpha' are computed by raising '10' to the sampled value power.

Usage

sample_lr_values_log_scale(n_grid_points = 10, lo = 1e-07, hi = 1)

Arguments

n_grid_points	integer, number of values to sample
lo	minimum learning rate value to return
hi	maximum learning rate value to return

Value

vector of 'alpha' samples within range [log(lo, 10), log(hi, 10)] Potentially run this after the lr_range_test to narrow?

ifrit98/R2deepR documentation built on June 19, 2020, 7:45 a.m.