Development version

You are reading the latest (development) docs. For stable documentation, see stable.

Stop Early When Fitness Plateaus

Time: 5 min | Difficulty: Beginner

What This Solves

Running all N generations wastes time when the search converges after generation 5. ConsecutiveStopping ends the search automatically when the best score doesn't improve for K consecutive generations.

Recipe

from sklearn.datasets import load_breast_cancer
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import StratifiedKFold, train_test_split

from sklearn_genetic import GASearchCV, EvolutionConfig, RuntimeConfig
from sklearn_genetic.callbacks import ConsecutiveStopping
from sklearn_genetic.space import Continuous, Integer

X, y = load_breast_cancer(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, stratify=y, random_state=42
)

ga = GASearchCV(
    estimator=RandomForestClassifier(random_state=42, n_jobs=-1),
    param_grid={
        "n_estimators": Integer(50, 300),
        "max_depth":    Integer(3, 20),
        "max_features": Continuous(0.1, 1.0),
    },
    scoring="roc_auc",
    cv=StratifiedKFold(n_splits=5, shuffle=True, random_state=42),
    evolution_config=EvolutionConfig(
        population_size=20,
        generations=50,    # upper limit — early stopping will terminate sooner
        elitism=True,
    ),
    runtime_config=RuntimeConfig(n_jobs=-1, verbose=True),
    random_state=42,
)

callback = ConsecutiveStopping(
    generations=5,          # stop if best score doesn't improve for 5 generations
    metric="fitness_best",  # track the running best (not generation mean)
)

ga.fit(X_train, y_train, callbacks=[callback])

print(f"Stopped after {ga.history[-1]['gen']} generations (max was 50)")
print(f"Best CV ROC AUC: {ga.best_score_:.4f}")

Available Metrics for ConsecutiveStopping

metric=	What it tracks
"fitness_best"	Best score seen so far (monotonically non-decreasing)
"fitness"	Current generation mean score
"fitness_std"	Standard deviation of current generation scores

Combine with a Time Budget

from sklearn_genetic.callbacks import ConsecutiveStopping, TimerStopping

ga.fit(
    X_train, y_train,
    callbacks=[
        ConsecutiveStopping(generations=5, metric="fitness_best"),
        TimerStopping(total_seconds=120),   # also stop if search takes > 2 min
    ]
)

Key Points

• generations=50 as upper limit: Set a generous max — early stopping will end it sooner in practice.
• "fitness_best" vs "fitness": "fitness_best" only increases; the search stops when the best-ever score stalls. "fitness" tracks the current generation mean, which can oscillate.
• Multiple callbacks: Both are checked after each generation. The search stops when any callback signals completion.

See Also

• Stop After a Time Budget — wall-clock limit
• Early Stopping and Callbacks — all available callbacks
• Writing Custom Callbacks — define your own stopping criterion