"""
# -*- coding: utf-8 -*-
#
# Copyright 2021 Michael Büsch <m@bues.ch>
#
# Licensed under the Apache License version 2.0
# or the MIT license, at your option.
# SPDX-License-Identifier: Apache-2.0 OR MIT
#
"""

from mlplib.activation import *
import numpy as np

def test_sigmoid():
    sigmoid = Sigmoid()
    assert np.allclose(sigmoid.fn(np.array([-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5])),
                       np.array([0.00669285, 0.01798621, 0.04742587, 0.11920292, 0.26894142, 0.5,
                                 0.73105858, 0.88079708, 0.95257413, 0.98201379, 0.99330715]))

def test_sigmoid_d():
    sigmoid = Sigmoid()
    assert np.allclose(sigmoid.fn_d(np.array([-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5])),
                       np.array([0.00664806, 0.01766271, 0.04517666, 0.10499359, 0.19661193, 0.25,
                                 0.19661193, 0.10499359, 0.04517666, 0.01766271, 0.00664806]))

def test_relu():
    relu = ReLU()
    assert np.all(relu.fn(np.array([1, 2, -1, -2])) == np.array([1, 2, 0, 0]))

def test_relu_d():
    relu = ReLU()
    assert np.all(relu.fn_d(np.array([1, 2, -1, -2])) == np.array([1, 1, 0, 0]))

def test_lrelu():
    lrelu = LReLU(alpha=0.1)
    assert np.all(lrelu.fn(np.array([1, 2, -1, -2])) == np.array([1, 2, -0.1, -0.2]))

def test_lrelu_d():
    lrelu = LReLU(alpha=0.1)
    assert np.all(lrelu.fn_d(np.array([1, 2, -1, -2])) == np.array([1, 1, 0.1, 0.1]))

def test_tanh():
    tanh = Tanh()
    assert np.allclose(tanh.fn(np.array([-3, -2, -1, -0.5, 0, 0.5, 1, 2, 3])),
                       np.array([-0.99505475, -0.96402758, -0.76159416, -0.46211716, 0,
                                 0.46211716, 0.76159416, 0.96402758, 0.99505475]))

def test_tanh_d():
    tanh = Tanh()
    assert np.allclose(tanh.fn_d(np.array([-3, -2, -1, -0.5, 0, 0.5, 1, 2, 3])),
                       np.array([0.00986604, 0.07065082, 0.41997434, 0.78644773, 1,
                                 0.78644773, 0.41997434, 0.07065082, 0.00986604]))

def test_softmax():
    # Test data Copyright 2008-2021, The SciPy community.
    inp = np.array([[1, 0.5, 0.2, 3],
                    [1, -1, 7, 3],
                    [2, 12, 13, 3]])
    out = Softmax().fn(inp)
    assert np.allclose(np.sum(out, axis=1),
                       np.array([1.0, 1.0, 1.0]))
    assert np.allclose(out,
                       np.array([[1.05877e-1, 6.42177e-2, 4.75736e-2, 7.82332e-1],
                                 [2.42746e-3, 3.28521e-4, 9.79307e-1, 1.79366e-2],
                                 [1.22094e-5, 2.68929e-1, 7.31025e-1, 3.31885e-5]]))

if __name__ == "__main__":
    import matplotlib.pyplot as plt
    fig, axes = plt.subplots(5, 2)
    fig.tight_layout()
    step = 0.01

    sigmoid = Sigmoid()
    ax = axes[0][0]
    ax.set_title("sigmoid")
    ax.set_ylim(-0.1, 1.1)
    x = np.arange(-5, 5, step)
    ax.plot(x, sigmoid.fn(x))
    ax = axes[0][1]
    ax.set_title("sigmoid derivative")
    ax.set_ylim(-0.1, 1.1)
    ax.plot(x, sigmoid.fn_d(x))

    relu = ReLU()
    ax = axes[1][0]
    ax.set_title("relu")
    ax.set_ylim(-0.1, 1.1)
    x = np.arange(-5, 5, step)
    ax.plot(x, relu.fn(x))
    ax = axes[1][1]
    ax.set_title("relu derivative")
    ax.set_ylim(-0.1, 1.1)
    ax.plot(x, relu.fn_d(x))

    lrelu = LReLU(0.1)
    ax = axes[2][0]
    ax.set_title("leaky relu")
    ax.set_ylim(-1.1, 1.1)
    x = np.arange(-5, 5, step)
    ax.plot(x, lrelu.fn(x))
    ax = axes[2][1]
    ax.set_title("leaky relu derivative")
    ax.set_ylim(-0.1, 1.1)
    ax.plot(x, lrelu.fn_d(x))

    tanh = Tanh()
    ax = axes[3][0]
    ax.set_title("tanh")
    ax.set_ylim(-1.1, 1.1)
    x = np.arange(-2, 2, step)
    ax.plot(x, tanh.fn(x))
    ax = axes[3][1]
    ax.set_title("tanh derivative")
    ax.set_ylim(-1.1, 1.1)
    ax.plot(x, tanh.fn_d(x))

    softmax = Softmax()
    ax = axes[4][0]
    ax.set_title("softmax")
    ax.set_ylim(-0.1, 1.1)
    x = np.array([[0.0, 0.1, 0.2, 0.4, 2.5, 0.4, 0.2, 0.1, 0.0]])
    xs = softmax.fn(x)
    ax.plot(xs.flatten())
    ax.plot(x.flatten())
    ax = axes[4][1]
    ax.set_title("softmax derivative")
    ax.set_ylim(-0.1, 1.1)
    xd = softmax.fn_d(x)
    ax.plot(xd.flatten())

    plt.show()

# vim: ts=4 sw=4 expandtab