Use splitfngr with lbfgs

Use lbfgs function from the lbfgs package but pass in a single function that returns both the function and gradient together in a list. Useful when the function and gradient are expensive to calculate and can be calculated faster together than separate.

Usage

lbfgs_share(fngr, vars, ...)

Arguments

fngr

A function that returns a list of two elements: the function value and the gradient value.

vars

Initial values for the parameters to be optimized over. Will be passed to lbfgs as vars argument.

...

Other arguments passed to lbfgs

Value

Result from running lbfgs on the given function

Examples

quad_share <- function(x){list(sum(x^4), 4*x^3)}
lbfgs_share(vars=c(3, -5), fngr=quad_share)
#> Iteration 1: 
#> fx = 322.314
#> 
#>   xnorm = 4.89476, gnorm = 274.43, step = 0.00195492
#> 
#> Iteration 2: 
#> fx = 100.763
#> 
#>   xnorm = 3.71359, gnorm = 111.211, step = 1
#> 
#> Iteration 3: 
#> fx = 34.1212
#> 
#>   xnorm = 2.84282, gnorm = 48.9431, step = 1
#> 
#> Iteration 4: 
#> fx = 10.9268
#> 
#>   xnorm = 2.14003, gnorm = 20.798, step = 1
#> 
#> Iteration 5: 
#> fx = 3.56744
#> 
#>   xnorm = 1.61783, gnorm = 8.98049, step = 1
#> 
#> Iteration 6: 
#> fx = 1.15599
#> 
#>   xnorm = 1.22065, gnorm = 3.85682, step = 1
#> 
#> Iteration 7: 
#> fx = 0.375674
#> 
#>   xnorm = 0.92163, gnorm = 1.66003, step = 1
#> 
#> Iteration 8: 
#> fx = 0.12195
#> 
#>   xnorm = 0.695664, gnorm = 0.713914, step = 1
#> 
#> Iteration 9: 
#> fx = 0.0396042
#> 
#>   xnorm = 0.525157, gnorm = 0.307125, step = 1
#> 
#> Iteration 10: 
#> fx = 0.0128596
#> 
#>   xnorm = 0.396425, gnorm = 0.132108, step = 1
#> 
#> Iteration 11: 
#> fx = 0.00417581
#> 
#>   xnorm = 0.299254, gnorm = 0.0568283, step = 1
#> 
#> Iteration 12: 
#> fx = 0.00135595
#> 
#>   xnorm = 0.225899, gnorm = 0.0244451, step = 1
#> 
#> Iteration 13: 
#> fx = 0.000440303
#> 
#>   xnorm = 0.170527, gnorm = 0.0105153, step = 1
#> 
#> Iteration 14: 
#> fx = 0.000142974
#> 
#>   xnorm = 0.128727, gnorm = 0.00452327, step = 1
#> 
#> Iteration 15: 
#> fx = 4.64263e-05
#> 
#>   xnorm = 0.0971729, gnorm = 0.00194573, step = 1
#> 
#> Iteration 16: 
#> fx = 1.50755e-05
#> 
#>   xnorm = 0.0733536, gnorm = 0.000836974, step = 1
#> 
#> Iteration 17: 
#> fx = 4.89527e-06
#> 
#>   xnorm = 0.055373, gnorm = 0.000360032, step = 1
#> 
#> Iteration 18: 
#> fx = 1.58958e-06
#> 
#>   xnorm = 0.0417999, gnorm = 0.000154871, step = 1
#> 
#> Iteration 19: 
#> fx = 5.16166e-07
#> 
#>   xnorm = 0.0315538, gnorm = 6.66195e-05, step = 1
#> 
#> Iteration 20: 
#> fx = 1.67608e-07
#> 
#>   xnorm = 0.0238192, gnorm = 2.8657e-05, step = 1
#> 
#> Iteration 21: 
#> fx = 5.44255e-08
#> 
#>   xnorm = 0.0179806, gnorm = 1.23271e-05, step = 1
#> 
#> Iteration 22: 
#> fx = 1.76729e-08
#> 
#>   xnorm = 0.0135732, gnorm = 5.30262e-06, step = 1
#> 
#> L-BFGS optimization terminated with status code = 0
#> fx = 1.76729e-08
#> 
#> $value
#> [1] 1.767293e-08
#> 
#> $par
#> [1]  0.008565892 -0.010528829
#> 
#> $convergence
#> [1] 0
#>