Exponential learning-rate decay (Geron Ch 11, morie.fn hmlrex)

eta_t = eta0 * decay^t, plus the number of steps per tenfold drop (Inf when decay = 1).

Usage

morie_geron_lr_exponential(eta0, decay, t)

Arguments

eta0

Positive initial rate.

decay

In (0, 1].

t

Step or vector of steps.

Value

List with eta, steps_per_decade, eta0, decay, estimate, n.

Examples

morie_geron_lr_exponential(0.1, 0.9, c(0, 1, 5, 10))
#> $eta
#> [1] 0.10000000 0.09000000 0.05904900 0.03486784
#> 
#> $steps_per_decade
#> [1] 21.85435
#> 
#> $eta0
#> [1] 0.1
#> 
#> $decay
#> [1] 0.9
#> 
#> $estimate
#> [1] 0.1
#> 
#> $n
#> [1] 4
#> 
#> $method
#> [1] "Exponential learning-rate decay"
#>