الشبكة العصبية من جوه

import numpy as np

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

#            weather  friends_free  have_money
inputs = np.array([0.9, 1.0, 0.2])
weights = np.array([1.5, 2.0, 1.0])      # how much each factor matters to you
bias = -2.5                              # you're lazy by default :)

z = inputs @ weights + bias              # the dot product from phase 3!
print(f"score = {z:.2f}")
print(f"probability of going out = {sigmoid(z):.0%}")

import numpy as np

rng = np.random.default_rng(0)
x = rng.normal(size=(5, 3))          # 5 samples, 3 features each

W1 = rng.normal(size=(3, 4))         # layer 1: 3 inputs -> 4 neurons
W2 = rng.normal(size=(4, 1))         # layer 2: 4 inputs -> 1 output

hidden = np.maximum(0, x @ W1)       # ReLU activation
output = hidden @ W2

print("input: ", x.shape)
print("hidden:", hidden.shape)
print("output:", output.shape)

import numpy as np
import matplotlib.pyplot as plt

z = np.linspace(-5, 5, 200)
fig, axes = plt.subplots(1, 3, figsize=(10, 2.8))
funcs = {
    "ReLU: max(0, z)": np.maximum(0, z),
    "Sigmoid: 0 to 1": 1 / (1 + np.exp(-z)),
    "Tanh: -1 to 1": np.tanh(z),
}
for ax, (name, y) in zip(axes, funcs.items()):
    ax.plot(z, y, color="#3776AB", linewidth=2.5)
    ax.axhline(0, color="gray", linewidth=0.5)
    ax.axvline(0, color="gray", linewidth=0.5)
    ax.set_title(name)
plt.tight_layout()
plt.show()