# -*- coding: utf-8 -*-

# Author: Vincent Dubourg <vincent.dubourg@gmail.com>
#         (mostly translation, see implementation details)
# License: BSD 3 clause

"""
The built-in regression models submodule for the gaussian_process module.
"""


import numpy as np


def constant(x):
    """
    Zero order polynomial (constant, p = 1) regression model.

    x --> f(x) = 1

    Parameters
    ----------
    x : array_like
        An array with shape (n_eval, n_features) giving the locations x at
        which the regression model should be evaluated.

    Returns
    -------
    f : array_like
        An array with shape (n_eval, p) with the values of the regression
        model.
    """
    x = np.asarray(x, dtype=np.float64)
    n_eval = x.shape[0]
    f = np.ones([n_eval, 1])
    return f


def linear(x):
    """
    First order polynomial (linear, p = n+1) regression model.

    x --> f(x) = [ 1, x_1, ..., x_n ].T

    Parameters
    ----------
    x : array_like
        An array with shape (n_eval, n_features) giving the locations x at
        which the regression model should be evaluated.

    Returns
    -------
    f : array_like
        An array with shape (n_eval, p) with the values of the regression
        model.
    """
    x = np.asarray(x, dtype=np.float64)
    n_eval = x.shape[0]
    f = np.hstack([np.ones([n_eval, 1]), x])
    return f


def quadratic(x):
    """
    Second order polynomial (quadratic, p = n*(n-1)/2+n+1) regression model.

    x --> f(x) = [ 1, { x_i, i = 1,...,n }, { x_i * x_j,  (i,j) = 1,...,n } ].T
                                                          i > j

    Parameters
    ----------
    x : array_like
        An array with shape (n_eval, n_features) giving the locations x at
        which the regression model should be evaluated.

    Returns
    -------
    f : array_like
        An array with shape (n_eval, p) with the values of the regression
        model.
    """

    x = np.asarray(x, dtype=np.float64)
    n_eval, n_features = x.shape
    f = np.hstack([np.ones([n_eval, 1]), x])
    for k in range(n_features):
        f = np.hstack([f, x[:, k, np.newaxis] * x[:, k:]])

    return f