A notebook showing how some very common activation functions (used for deep-learning for example) look like. Enjoy!



In [20]:

    
%matplotlib inline
import numpy as np
import matplotlib.pylab as plt
import seaborn as sns



In [16]:

    
sns.set(font_scale=1.5)

Define the activiation metrics



In [39]:

    
def sigmoid(x):
    return 1 / (1 + np.exp(-x))



In [8]:

    
def tanh(x):
    return np.tanh(x)



In [13]:

    
def relu(x):
    return np.maximum(x, 0) # element-wise maximum

Plot them



In [52]:

    
class ActivationPlots(object):
    def __init__(self, metrics):
        self.x = np.arange(-10, 10, 0.1)
        self.metrics = metrics
        self.n_plots = len(self.metrics)
    def build(self, axes):
        for ax, metric in zip(axes, self.metrics):
            y = metric(self.x)
            ax.plot(self.x, y)
            ax.set_title(str(metric.__name__))
        return axes
    def plot(self):
        n_rows = self.n_plots % 2
        n_cols = int(self.n_plots / n_rows)
        fig, axes = plt.subplots(n_rows, n_cols, figsize=(12, 4))
        self.build(axes)



In [53]:

    
ActivationPlots([sigmoid, tanh, relu]).plot()