6 · Splitting, Validation & Experiment Design · Cross-Validation & Data Splitting

Nested cross-validation

Nested CV separates hyperparameter selection from performance estimation: an inner loop tunes the model and an outer loop estimates generalisation. It is computationally expensive but reduces optimisation bias. The important practical question is not only how the technique is defined, but what assumptions it introduces, which data are allowed to influence it, and how its effect should be validated on unseen evidence.

Reference lessonPython exampleVisual explanation

Intuition first

What this concept means in practice

The practical value of Nested cross-validation comes from understanding both the transformation and the boundary around it: what information is allowed to enter, what assumption is being made, and how you know the result is still valid after the transformation.

A beginner-friendly way to reason about it is to start with a tiny case where the correct result can be checked independently. Once the mechanism is clear, scale the exact same reasoning to larger tables, pipelines or models.

PurposeUse when datasets are limited and tuning is extensive enough to bias a single validation estimate.

MechanismUse an inner CV loop to choose hyperparameters for each outer-training fold, then score the selected pipeline once on that outer test fold.

EvidenceInspect intermediate and final output; compare with an independent expectation.

Main cautionThe final reported score must come from outer folds, not the best inner-CV score.

Mechanism

Trace the operation from input to decision

Use an inner CV loop to choose hyperparameters for each outer-training fold, then score the selected pipeline once on that outer test fold.

1Input→

2Apply rule→

3Inspect state→

4Validate→

5Use result

Key rule

outer evaluation ⟂ inner tuning

Visual explanation

Make the structure visible

The interactive view uses a concept-specific plot when the topic maps naturally to one; otherwise it uses a workflow view instead of leaving a broken placeholder.

Loading visual…

Practical example

Where you would use it

Five outer folds each run a three-fold inner Bayesian search; outer scores estimate generalisation after tuning.

Use when

Use when datasets are limited and tuning is extensive enough to bias a single validation estimate.

Pitfall

What can make the result misleading

Watch out

The final reported score must come from outer folds, not the best inner-CV score.

A useful diagnostic question is: Could the same code still run successfully if the analytical assumption were wrong? If yes, add an explicit validation check rather than relying on execution success.

Implementation

Miniature Python example

Keep the example small enough that you can inspect each stage manually.

Python

# Purpose: demonstrate Nested cross-validation with a small, inspectable example.
# Follow the comments and printed stages to connect each operation with its result.
# Import the library or helper used in this example.
# Step 1 — Import the module so its functions/classes are available to the rest of this example.
import numpy as np
# Import the library or helper used in this example.
# Step 2 — Import only the named objects needed by the following steps, keeping dependencies explicit.
from sklearn.model_selection import StratifiedKFold

# Create the numerical values used in the calculation.
# Step 3 — Compute the right-hand expression and store its result in `X` for the next step.
X = np.arange(48).reshape(24,2)
# Step 4 — Construct `y` as an array so vectorised numerical operations can be applied consistently.
y = np.array([0]*12 + [1]*12)
# Store this intermediate value with a descriptive name for the next step.
# Step 5 — Compute the right-hand expression and store its result in `cv` for the next step.
cv = StratifiedKFold(n_splits=4, shuffle=True, random_state=7)
# Print this intermediate result so you can verify the workflow step by step.
# Step 6 — Display the current value explicitly so the result/state can be inspected during execution.
print("STEP 1 · Samples:", len(y), "class counts:", np.bincount(y).tolist())
# Iterate through the current values one item or step at a time.
# Step 7 — Iterate through the collection so the indented block is applied once for each item.
for fold,(tr,va) in enumerate(cv.split(X,y),1):
    # Print this intermediate result so you can verify the workflow step by step.
    # Step 8 — Display the current value explicitly so the result/state can be inspected during execution.
    print(f"STEP 2 · Fold {fold}: train={len(tr)} valid={len(va)} valid_classes={np.bincount(y[va]).tolist()}")
# Print this intermediate result so you can verify the workflow step by step.
# Step 9 — Display the current value explicitly so the result/state can be inspected during execution.
print("STEP 3 · Each sample is validation data once across the four folds.")

Expected / illustrative output

STEP 1 · Samples: 24 class counts: [12, 12]
STEP 2 · Fold 1: train=18 valid=6 valid_classes=[3, 3]
STEP 2 · Fold 2: train=18 valid=6 valid_classes=[3, 3]
STEP 2 · Fold 3: train=18 valid=6 valid_classes=[3, 3]
STEP 2 · Fold 4: train=18 valid=6 valid_classes=[3, 3]
STEP 3 · Each sample is validation data once across the four folds.

Implementation checklist

Before you move on

Can you state what data or object enters the operation?
Can you explain what changes and what must remain invariant?
Have you checked the result on a tiny case you can verify independently?
Have you considered the main failure mode described above?
Can the operation be reproduced from code/formulas and documented assumptions?