Little Bits of Artificial Intelligence

1. Simple Neural Network (2-1-1 Architecture)

• 2 inputs: Sweetness & Moistness
• 1 hidden layer with 2 neurons
• 1 second hidden layer with 1 neuron
• 1 output neuron (final score)

# Input: Cake features
x1 = 7   # Sweetness
x2 = 6   # Moistness

# -------- Hidden Layer 1 (2 neurons) ----------
# Weights and biases
w11, w21, b1 = 0.5, 0.4, 1    # Neuron 1
w12, w22, b2 = -0.3, 0.8, 0.5 # Neuron 2

# ReLU Activation Function
def relu(x):
    return max(0, x)

# Neuron 1 Output
z1 = x1 * w11 + x2 * w21 + b1
a1 = relu(z1)

# Neuron 2 Output
z2 = x1 * w12 + x2 * w22 + b2
a2 = relu(z2)

# -------- Hidden Layer 2 (1 neuron) ----------
w31, w32, b3 = 0.6, 0.9, -1

z3 = a1 * w31 + a2 * w32 + b3
a3 = relu(z3)

# -------- Output Layer ----------
w4, b4 = 1.2, 0.3

final_score = a3 * w4 + b4

# -------- Print Output --------
print("Final Cake Score:", round(final_score, 3))

Output:
Final Cake Score: 7.524

What This Simulates:

• Hidden Layer 1: Experts analyzing raw features.
• Hidden Layer 2: Combines those mini-judgments.
• Output Layer: Final score, like a competition result.

Each step shows how the layered thinking (ensemble effect) builds up — from raw cake traits to a refined evaluation.