Conditionals and Branching

Machine Learning is often about making decisions. Conditionals allow our code to react differently depending on the input. Whether it's checking if a dataset is empty or deciding which model to load, if-else logic is the foundation of programmatic decision-making.

1. The if, elif, and else Structure

Python uses indentation to define the scope of conditional blocks.

accuracy = 0.85

if accuracy > 0.90:
    print("Excellent model performance!")
elif accuracy > 0.70:
    print("Good performance, but could be improved.")
else:
    print("Model needs retraining.")

2. Comparison and Logical Operators

Conditionals rely on boolean expressions that evaluate to either True or False.

A. Comparison Operators

• == (Equal to)
• != (Not equal to)
• > / < (Greater/Less than)
• >= / <= (Greater/Less than or equal to)

B. Logical Operators (Chaining)

• and: Both conditions must be True.
• or: At least one condition must be True.
• not: Reverses the boolean value.

# Check if learning rate is within a safe range
lr = 0.001
if lr > 0 and lr < 0.1:
    print("Learning rate is valid.")

3. The "ReLU" Example: Math meets Logic

One of the most famous conditional operations in Deep Learning is the Rectified Linear Unit (ReLU) activation function.

$$\text{ReLU}(x) = \max(0, x)$$

In Python code, this is a simple conditional:

def relu(x):
    if x > 0:
        return x
    else:
        return 0

4. Truthiness and Identity

In ML data cleaning, we often check if a variable actually contains data.

• Falsy values: None, 0, 0.0, "" (empty string), [] (empty list), {} (empty dict).
• Truthy values: Everything else.

features = get_features()

if not features:
    print("Warning: No features found in dataset!")

is vs ==

• == checks for Value equality (Are the numbers the same?).
• is checks for Identity (Are they the exact same object in memory?).

5. Inline Conditionals (Ternary Operator)

For simple logic, Python allows a one-liner known as a ternary operator.

# status = "Spam" if probability > 0.5 else "Not Spam"
prediction = "Positive" if y_hat > 0.5 else "Negative"

6. Match-Case (Python 3.10+)

For complex branching based on specific patterns (like different file extensions or model types), the match statement provides a cleaner syntax than multiple elif blocks.

optimizer_type = "adam"

match optimizer_type:
    case "sgd":
        print("Using Stochastic Gradient Descent")
    case "adam":
        print("Using Adam Optimizer")
    case _:
        print("Using Default Optimizer")

---

Decision-making is key, but to keep our code clean, we shouldn't repeat our logic. We need to wrap our conditionals and loops into reusable components.