Skip to main content

3.3 Neuron Layer

A layer is simply multiple neurons calculating their outputs together.

x₁ ──┐
x₂ ──┼─→ Neuron 1 (Financial Strength) ──→ y₁ ─┐
x₃ ──┘                                         │
					       ├──→ Neuron 3 (loan decision) ──→output
x₁ ──┐                                         │
x₂ ──┼──→ Neuron 2 (Risk) ──────────────→ y₂ ──┘
x₃ ──┘

Lets take real world scenerio for this
x₁ = 2000 sq ft
x₂ = 3 bedrooms
x₃ = 2 bathrooms

and although neurons are auto trained and we can never know what they mean but lets say after train they behave like this
Neuron 1 -"size/value" neuron Neuron 2 -"comfort" neuron

Then the final neuron receives those two learned signals: Neuron 3 - Loan decision.

Here is how such a neural network be made​

import numpy as np

# Inputs
x = np.array([2, 3, 4])
print('inputs are :1000 sqft, bedrooms, bathrooms : ',x)
# First layer: 2 neurons
W1 = np.array([
    [1, 2, 3],
    [2, 1, 1]
])

b1 = np.array([5, 2])

# Outputs of first layer
y1 = W1 @ x + b1

print("First layer:", y1)


# Second layer: 1 neuron
W2 = np.array([2, 3])
b2 = 1

# Final output
y2 = W2 @ y1 + b2

print("Final output:", y2)