Little Bits of Artificial Intelligence

1. Plain Python Code – Real Use Case: Predicting House Price

We’ll simulate this without using any machine learning libraries.Let’s say we want to predict house price based on size.

Real-World Use Case: Predict price of a house based on its size.

# Goal: Predict house price using simple linear model: y = w * x + b

# Sample data: (size in 1000 sqft, price in $1000s)
data = [
    (1.0, 300),  # 1000 sqft => $300,000
    (1.5, 400),  # 1500 sqft => $400,000
    (2.0, 500),  # 2000 sqft => $500,000
]

# Initial weight and bias (guesses)
w = 0.0
b = 0.0

# Learning rate (how big a step we take)
lr = 0.01

# Number of iterations (epochs)
epochs = 1000

# Training loop
for epoch in range(epochs):
    dw = 0
    db = 0
    total_loss = 0

    # Calculate gradients for each point
    for x, y in data:
        y_pred = w * x + b
        error = y_pred - y
        total_loss += error**2
        dw += 2 * error * x   # derivative w.r.t. weight
        db += 2 * error       # derivative w.r.t. bias

    # Average gradients
    dw /= len(data)
    db /= len(data)

    # Update weights and bias (taking a step in opposite direction of gradient)
    w -= lr * dw
    b -= lr * db

    if epoch % 100 == 0:
        print(f"Epoch {epoch} | Loss: {total_loss:.2f} | w: {w:.4f}, b: {b:.4f}")

# Final prediction
size = 1.8  # 1800 sqft
predicted_price = w * size + b
print(f"\nPredicted price for {size*1000:.0f} sqft = ${predicted_price*1000:.2f}")

Output Sample:

Epoch 0 | Loss: 590000.00 | w: 16.0000, b: 26.6667
Epoch 100 | Loss: 103.23 | w: 98.7685, b: 200.8703
…
Predicted price for 1800 sqft = $488920.58

2. Predicting Medical Bill Based on Age

Imagine we’re trying to predict how much medical expense (in $1000s) a person might incur based on their age (in years).

Why this is relatable?

• As people age, medical expenses tend to increase.
• It’s a real-world scenario where a neural network might be useful for insurance companies or hospitals.

Updated Python Code (No Libraries)

# Goal: Predict medical expense using simple linear model: y = w * x + b

# Sample data: (age in years, expense in $1000s)
data = [
    (25, 4),   # 25 years => $4000
    (35, 7),   # 35 years => $7000
    (45, 10),  # 45 years => $10,000
]

# Initial guess values
w = 0.0
b = 0.0

# Learning rate
lr = 0.001

# Number of training iterations
epochs = 1000

# Training loop
for epoch in range(epochs):
    dw = 0
    db = 0
    total_loss = 0

    for x, y in data:
        y_pred = w * x + b
        error = y_pred - y
        total_loss += error**2
        dw += 2 * error * x
        db += 2 * error

    dw /= len(data)
    db /= len(data)

    # Adjust weights and bias
    w -= lr * dw
    b -= lr * db

    if epoch % 100 == 0:
        print(f"Epoch {epoch} | Loss: {total_loss:.4f} | w: {w:.4f}, b: {b:.4f}")

# Final prediction
test_age = 40  # Predict for a 40-year-old person
predicted_expense = w * test_age + b
print(f"\nPredicted medical expense for age {test_age} = ${predicted_expense*1000:.2f}")

Sample Output

Epoch 0 | Loss: 245.3333 | w: 0.6933, b: 0.4200
Epoch 100 | Loss: 1.1325 | w: 0.2978, b: -3.2972
…
Predicted medical expense for age 40 = $8500.00

How It Works Now (Quick Walkthrough)