lab_utils_softmax.py

import numpy as np
import matplotlib.pyplot as plt
plt.style.use('./deeplearning.mplstyle')
import tensorflow as tf
from IPython.display import display, Markdown, Latex
from matplotlib.widgets import Slider
from lab_utils_common import dlc


def plt_softmax(my_softmax):
    fig, ax = plt.subplots(1,2,figsize=(8,4))
    plt.subplots_adjust(bottom=0.35)

    axz0 = fig.add_axes([0.15, 0.10, 0.30, 0.03]) # [left, bottom, width, height]
    axz1 = fig.add_axes([0.15, 0.15, 0.30, 0.03])
    axz2 = fig.add_axes([0.15, 0.20, 0.30, 0.03])
    axz3 = fig.add_axes([0.15, 0.25, 0.30, 0.03])

    z3 = Slider(axz3, 'z3', 0.1, 10.0, valinit=4, valstep=0.1)
    z2 = Slider(axz2, 'z2', 0.1, 10.0, valinit=3, valstep=0.1)
    z1 = Slider(axz1, 'z1', 0.1, 10.0, valinit=2, valstep=0.1)
    z0 = Slider(axz0, 'z0', 0.1, 10.0, valinit=1, valstep=0.1)

    z = np.array(['z0','z1','z2','z3'])
    bar = ax[0].barh(z, height=0.6, width=[z0.val,z1.val,z2.val,z3.val], left=None, align='center')
    bars = bar.get_children()
    ax[0].set_xlim([0,10])
    ax[0].set_title("z input to softmax")

    a = my_softmax(np.array([z0.val,z1.val,z2.val,z3.val]))
    anames = np.array(['a0','a1','a2','a3'])
    sbar = ax[1].barh(anames, height=0.6, width=a, left=None, align='center',color=dlc["dldarkred"])
    sbars = sbar.get_children()
    ax[1].set_xlim([0,1])
    ax[1].set_title("softmax(z)")

    def update(val):
        bars[0].set_width(z0.val)
        bars[1].set_width(z1.val)
        bars[2].set_width(z2.val)
        bars[3].set_width(z3.val)
        a = my_softmax(np.array([z0.val,z1.val,z2.val,z3.val]))
        sbars[0].set_width(a[0])
        sbars[1].set_width(a[1])
        sbars[2].set_width(a[2])
        sbars[3].set_width(a[3])

        fig.canvas.draw_idle()

    z0.on_changed(update)
    z1.on_changed(update)
    z2.on_changed(update)
    z3.on_changed(update)

    plt.show()