Module Fit.Make

Parameters

module P : sig ... end

Signature

type prm

abstract type of parameters

val n : int

number of neurons

val n_prm : int

total number of parameters

val pack : (float * Filter.t * Filter.t array option) array -> prm

packs an array of (baseline, filter, couplings) into a prm type

val unpack : prm -> (float * Filter.t * Filter.t array option) array

unpacks a prm into an array of (baseline, filter, couplings)

val log_likelihood : prm -> float

evaluates the log likelihood for a given set of parameters

val loss : prm -> float

evaluates the loss for a given set of parametesr

val fit : ?check_gradient:bool -> ?in_dir:(string -> string) -> int -> prm

fit max_iter fits the model using BFGS; if ~in_dir is specified, the decreasing loss function is saved on the fly in in_dir "loss"