-module(gleastsq).
-compile([no_auto_import, nowarn_unused_vars, nowarn_unused_function, nowarn_nomatch, inline]).
-define(FILEPATH, "src\\gleastsq.gleam").
-export([least_squares/5, levenberg_marquardt/5, gauss_newton/5, trust_region_reflective/7]).

-if(?OTP_RELEASE >= 27).
-define(MODULEDOC(Str), -moduledoc(Str)).
-define(DOC(Str), -doc(Str)).
-else.
-define(MODULEDOC(Str), -compile([])).
-define(DOC(Str), -compile([])).
-endif.

-file("src\\gleastsq.gleam", 10).
?DOC(
    " The `least_squares` function is an alias for the `levenberg_marquardt` function.\n"
    " Check the documentation of the `levenberg_marquardt` function for more information.\n"
).
-spec least_squares(
    list(float()),
    list(float()),
    fun((float(), list(float())) -> float()),
    list(float()),
    list(gleastsq@options:least_square_options())
) -> {ok, list(float())} | {error, gleastsq@errors:fit_errors()}.
least_squares(X, Y, Func, Initial_params, Opts) ->
    gleastsq@internal@methods@levenberg_marquardt:levenberg_marquardt(
        X,
        Y,
        Func,
        Initial_params,
        gleastsq@internal@params:decode_params(Opts)
    ).

-file("src\\gleastsq.gleam", 76).
?DOC(
    " The `levenberg_marquardt` function performs the Levenberg-Marquardt optimization algorithm.\n"
    " It is used to solve non-linear least squares problems. This function takes as input the data points,\n"
    " the model function, and several optional parameters to control the optimization process.\n"
    "\n"
    " # Parameters\n"
    " - `x` (List(Float))\n"
    "     A list of x-values of the data points.\n"
    " - `y` (List(Float))\n"
    "     A list of y-values of the data points.\n"
    " - `func` (fn(Float, List(Float)) -> Float)\n"
    "     The model function that takes an x-value and a list of parameters, and returns the corresponding y-value.\n"
    " - `initial_params` (List(Float))\n"
    "     A list of initial guesses for the parameters of the model function.\n"
    "     This list must not be empty.\n"
    " - `opts` (List(LeastSquareOptions))\n"
    "     A list of optional parameters to control the optimization process.\n"
    "     The available options are:\n"
    "     - `Iterations(Int)`: The maximum number of iterations to perform. Default is 100.\n"
    "     - `Epsilon(Float)`: A small value to change x when calculating the derivatives for the function. Default is 0.0001.\n"
    "     - `Tolerance(Float)`: The convergence tolerance. Default is 0.0001.\n"
    "     - `Damping(Float)`: The initial value of the damping parameter. Default is 0.0001.\n"
    "     - `DampingIncrease(Float)`: The factor by which the damping parameter is increased when a step fails. Default is 10.0.\n"
    "     - `DampingDecrease(Float)`: The factor by which the damping parameter is decreased when a step succeeds. Default is 0.1.\n"
    "\n"
    " # Errors\n"
    " Returns `Error(WrongParameters(_))` if `x` and `y` have different lengths or\n"
    " if `initial_params` is empty.\n"
    "\n"
    " # Example\n"
    " ```gleam\n"
    " import gleam/io\n"
    " import gleastsq\n"
    " import gleastsq/options.{Iterations, Tolerance}\n"
    "\n"
    " fn parabola(x: Float, params: List(Float)) -> Float {\n"
    "   let assert [a, b, c] = params\n"
    "   a *. x *. x +. b *. x +. c\n"
    " }\n"
    "\n"
    " pub fn main() {\n"
    "   let x = [0.0, 1.0, 2.0, 3.0, 4.0, 5.0]\n"
    "   let y = [0.0, 1.0, 4.0, 9.0, 16.0, 25.0]\n"
    "   let initial_guess = [1.0, 1.0, 1.0]\n"
    "\n"
    "   let assert Ok(result) =\n"
    "     gleastsq.levenberg_marquardt(\n"
    "       x,\n"
    "       y,\n"
    "       parabola,\n"
    "       initial_guess,\n"
    "       opts: [Iterations(1000), Tolerance(0.001)]\n"
    "     )\n"
    "\n"
    "   io.debug(result) // [1.0, 0.0, 0.0] (within numerical error)\n"
    " }\n"
    " ```\n"
).
-spec levenberg_marquardt(
    list(float()),
    list(float()),
    fun((float(), list(float())) -> float()),
    list(float()),
    list(gleastsq@options:least_square_options())
) -> {ok, list(float())} | {error, gleastsq@errors:fit_errors()}.
levenberg_marquardt(X, Y, Func, Initial_params, Opts) ->
    gleastsq@internal@methods@levenberg_marquardt:levenberg_marquardt(
        X,
        Y,
        Func,
        Initial_params,
        gleastsq@internal@params:decode_params(Opts)
    ).

-file("src\\gleastsq.gleam", 140).
?DOC(
    " The `gauss_newton` function performs a basic least squares optimization algorithm.\n"
    " It is used to find the best-fit parameters for a given model function to a set of data points.\n"
    " This function takes as input the data points, the model function, and several optional parameters to control the optimization process.\n"
    "\n"
    " # Parameters\n"
    " - `x` (List(Float))\n"
    "     A list of x-values of the data points.\n"
    " - `y` (List(Float))\n"
    "     A list of y-values of the data points.\n"
    " - `func` (fn(Float, List(Float)) -> Float)\n"
    "     The model function that takes an x-value and a list of parameters, and returns the corresponding y-value.\n"
    " - `initial_params` (List(Float))\n"
    "     A list of initial guesses for the parameters of the model function.\n"
    "     This list must not be empty.\n"
    " - `opts` (List(LeastSquareOptions))\n"
    "     A list of optional parameters to control the optimization process.\n"
    "     The available options are:\n"
    "     - `Iterations(Int)`: The maximum number of iterations to perform. Default is 100.\n"
    "     - `Epsilon(Float)`: A small value to change x when calculating the derivatives for the function. Default is 0.0001.\n"
    "     - `Tolerance(Float)`: The convergence tolerance. Default is 0.0001.\n"
    "     - `Damping(Float)`: The value of the damping parameter. Default is 0.001.\n"
    "\n"
    " # Errors\n"
    " Returns `Error(WrongParameters(_))` if `x` and `y` have different lengths or\n"
    " if `initial_params` is empty.\n"
    "\n"
    " # Example\n"
    " ```gleam\n"
    " import gleam/io\n"
    " import gleastsq\n"
    " import gleastsq/options.{Iterations, Tolerance}\n"
    "\n"
    " fn parabola(x: Float, params: List(Float)) -> Float {\n"
    "   let assert [a, b, c] = params\n"
    "   a *. x *. x +. b *. x +. c\n"
    " }\n"
    "\n"
    " pub fn main() {\n"
    "   let x = [0.0, 1.0, 2.0, 3.0, 4.0, 5.0]\n"
    "   let y = [0.0, 1.0, 4.0, 9.0, 16.0, 25.0]\n"
    "   let initial_guess = [1.0, 1.0, 1.0]\n"
    "\n"
    "   let assert Ok(result) =\n"
    "     gleastsq.gauss_newton(\n"
    "       x,\n"
    "       y,\n"
    "       parabola,\n"
    "       initial_guess,\n"
    "       opts: [Iterations(1000), Tolerance(0.001)]\n"
    "     )\n"
    "\n"
    "   io.debug(result) // [1.0, 0.0, 0.0] (within numerical error)\n"
    " }\n"
    " ```\n"
).
-spec gauss_newton(
    list(float()),
    list(float()),
    fun((float(), list(float())) -> float()),
    list(float()),
    list(gleastsq@options:least_square_options())
) -> {ok, list(float())} | {error, gleastsq@errors:fit_errors()}.
gauss_newton(X, Y, Func, Initial_params, Opts) ->
    gleastsq@internal@methods@gauss_newton:gauss_newton(
        X,
        Y,
        Func,
        Initial_params,
        gleastsq@internal@params:decode_params(Opts)
    ).

-file("src\\gleastsq.gleam", 216).
?DOC(
    " The `trust_region_reflective` function performs a bounded trust-region least\n"
    " squares optimization.\n"
    " It is used to find the best-fit parameters for a given model function to a\n"
    " set of data points while respecting optional upper and lower bounds.\n"
    " This function takes as input the data points, the model function, and\n"
    " several optional parameters to control the optimization process.\n"
    "\n"
    " # Parameters\n"
    " - `x` (List(Float))\n"
    "     A list of x-values of the data points.\n"
    " - `y` (List(Float))\n"
    "     A list of y-values of the data points.\n"
    " - `func` (fn(Float, List(Float)) -> Float)\n"
    "     The model function that takes an x-value and a list of parameters, and returns the corresponding y-value.\n"
    " - `initial_params` (List(Float))\n"
    "     A list of initial guesses for the parameters of the model function.\n"
    "     This list must not be empty. Values outside the provided bounds are\n"
    "     clipped to the bounds before optimization begins.\n"
    " - `lower_bounds` (Option(List(Float)))\n"
    "     A list of lower bounds for the parameters of the model function. If the lower bound is `None`, then the parameter is unbounded from below.\n"
    " - `upper_bounds` (Option(List(Float)))\n"
    "     A list of upper bounds for the parameters of the model function. If the upper bound is `None`, then the parameter is unbounded from above.\n"
    " - `opts` (List(LeastSquareOptions))\n"
    "     A list of optional parameters to control the optimization process.\n"
    "     The available options are:\n"
    "     - `Iterations(Int)`: The maximum number of iterations to perform. Default is 100.\n"
    "     - `Epsilon(Float)`: A small value to change x when calculating the derivatives for the function. Default is 0.0001.\n"
    "     - `Tolerance(Float)`: The convergence tolerance. Default is 0.00001.\n"
    "     - `Damping(Float)`: The value of the damping parameter. Default is 0.001.\n"
    "\n"
    " # Errors\n"
    " Returns `Error(WrongParameters(_))` if `x` and `y` have different lengths,\n"
    " if `initial_params` is empty, or if the bound lists have different lengths\n"
    " from `initial_params`.\n"
    "\n"
    " # Example\n"
    " ```gleam\n"
    " import gleam/io\n"
    " import gleam/option.{Some}\n"
    " import gleastsq\n"
    " import gleastsq/options.{Iterations, Tolerance}\n"
    "\n"
    " fn parabola(x: Float, params: List(Float)) -> Float {\n"
    "   let assert [a, b, c] = params\n"
    "   a *. x *. x +. b *. x +. c\n"
    " }\n"
    "\n"
    " pub fn main() {\n"
    "   let x = [0.0, 1.0, 2.0, 3.0, 4.0, 5.0]\n"
    "   let y = [0.0, 1.0, 4.0, 9.0, 16.0, 25.0]\n"
    "   let initial_guess = [1.0, 1.0, 1.0]\n"
    "\n"
    "   let assert Ok(result) =\n"
    "     gleastsq.trust_region_reflective(\n"
    "       x,\n"
    "       y,\n"
    "       parabola,\n"
    "       initial_guess,\n"
    "       lower_bounds: Some([0.0, 0.0, 0.0]),\n"
    "       upper_bounds: Some([2.0, 2.0, 2.0]),\n"
    "       opts: [Iterations(1000), Tolerance(0.001)]\n"
    "     )\n"
    "\n"
    "   io.debug(result) // [1.0, 0.0, 0.0] (within numerical error)\n"
    " }\n"
    " ```\n"
).
-spec trust_region_reflective(
    list(float()),
    list(float()),
    fun((float(), list(float())) -> float()),
    list(float()),
    gleam@option:option(list(float())),
    gleam@option:option(list(float())),
    list(gleastsq@options:least_square_options())
) -> {ok, list(float())} | {error, gleastsq@errors:fit_errors()}.
trust_region_reflective(
    X,
    Y,
    Func,
    Initial_params,
    Lower_bounds,
    Upper_bounds,
    Opts
) ->
    gleastsq@internal@methods@trust_region_reflective:trust_region_reflective(
        X,
        Y,
        Func,
        Initial_params,
        Lower_bounds,
        Upper_bounds,
        gleastsq@internal@params:decode_params(Opts)
    ).