plotter.py

# -*- coding: utf-8 -*-
#!/usr/bin/env python3

__author__ = "Ilya Shoshin (Galarius)"

import numpy as np
from pylab import plt


def sigmoid(s, a=1):
    return 1 / (1 + np.exp(-a * s))


def sigmoidPrime(s):
    return s * (1 - s)


def sigmoidTanh(s, a=1):
    return np.tanh(s / a)


def plotSigmoidPrime(fname=None):
    _, ax = plt.subplots()
    xs = np.linspace(0.0, 1.0, num=50, endpoint=True)
    ys = [sigmoidPrime(x) for x in xs]
    ax.plot(xs, ys, "black")
    plt.title("y=sigmoid'(s)")
    plt.grid(True)
    if fname:
        plt.savefig(fname)
    plt.show()


def plotSigmoidExp(fname=None):
    _, ax = plt.subplots()
    xs = np.linspace(-10.0, 10.0, num=50, endpoint=True)
    ys = [sigmoid(x, 0.9) for x in xs]
    ax.plot(xs, ys, "black")
    plt.title("y=sigmoid(s)")
    plt.grid(True)
    if fname:
        plt.savefig(fname)
    plt.show()


def plotSigmoidTanh(fname=None):
    _, ax = plt.subplots()
    xs = np.linspace(-10.0, 10.0, num=50, endpoint=True)
    ys = [sigmoidTanh(x, 0.9) for x in xs]
    ax.plot(xs, ys, "black")
    plt.title("y=sigmoid(s)")
    plt.grid(True)
    if fname:
        plt.savefig(fname)
    plt.show()


def plotSigmoid(fname=None):
    _, ax = plt.subplots()
    xs = np.linspace(-10.0, 10.0, num=50, endpoint=True)
    ys = [sigmoid(0.5 * x) for x in xs]
    ax.plot(xs, ys, "black", linestyle="-", label="sig(0.5 * x)")
    ys = [sigmoid(1.0 * x) for x in xs]
    ax.plot(xs, ys, "black", linestyle="--", label="sig(1.0 * x)")
    ys = [sigmoid(2.5 * x) for x in xs]
    ax.plot(xs, ys, "black", linestyle="-.", label="sig(2.5 * x)")
    plt.title("y=sig(s * w1)")
    plt.grid(True)
    if fname:
        plt.savefig(fname)
    plt.show()


def plotSigmoidBias(fname=None):
    _, ax = plt.subplots()
    xs = np.linspace(-10.0, 10.0, num=50, endpoint=True)
    ys = [sigmoid(1.0 * x - 5.0) for x in xs]
    ax.plot(xs, ys, "black", linestyle="-", label="sig(1.0 * x - 1.0 * 5)")
    ys = [sigmoid(1.0 * x) for x in xs]
    ax.plot(xs, ys, "black", linestyle="--", label="sig(1.0 * x + 1.0 * 0)")
    ys = [sigmoid(1.0 * x + 5.0) for x in xs]
    ax.plot(xs, ys, "black", linestyle="-.", label="sig(1.0 * x + 1.0 * 5)")
    plt.title("y=sig(s * w1 + 1.0 * w2)")
    plt.grid(True)
    if fname:
        plt.savefig(fname)
    plt.show()


plotSigmoidPrime("plt_sig_e_prime.png")
plotSigmoidExp("plt_sig_e.png")
plotSigmoidTanh("plt_sig_tanh.png")
plotSigmoid("plt_sig_w1.png")
plotSigmoidBias("plt_sig_w1w2.png")