Neural Network (MLP) Optimization

Hyperparameter optimization for Multi-Layer Perceptron neural networks.

from sklearn.datasets import load_breast_cancer
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, f1_score
from sklearn.preprocessing import StandardScaler
import numpy as np
import plotly
from mloptimizer.interfaces import HyperparameterSpaceBuilder, GeneticSearch
from mloptimizer.application.reporting.plots import plotly_search_space, plotly_logbook

Load and prepare the dataset

print("Loading Breast Cancer dataset...")
data = load_breast_cancer()
X, y = data.data, data.target

# Scale features for neural networks
scaler = StandardScaler()
X = scaler.fit_transform(X)

print(f"Dataset shape: {X.shape}")

Loading Breast Cancer dataset...
Dataset shape: (569, 30)

Split the data

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

Define the hyperparameter space

hyperparam_space = HyperparameterSpaceBuilder.get_default_space(
    estimator_class=MLPClassifier
)

Configure and run the genetic optimization

genetic_params = {
    'generations': 5,
    'population_size': 8,
    'n_elites': 2,
    'seed': 42,
    'use_mlflow': False,
    'use_parallel': False
}

opt = GeneticSearch(
    estimator_class=MLPClassifier,
    hyperparam_space=hyperparam_space,
    cv=3,
    scoring='accuracy',
    **genetic_params
)

print("Starting MLPClassifier optimization...")
opt.fit(X_train, y_train)

/home/docs/checkouts/readthedocs.org/user_builds/mloptimizer/checkouts/master/examples/plot_mlp_neural_network.py:52: UserWarning: Expected mutations per offspring is very low (0.48). With mutpb=0.8, indpb=0.2, and 3 hyperparameters, the population will converge prematurely. Recommended: mutpb >= 0.8, indpb >= 0.2 (gives ~0.5 mutations/offspring).
  opt = GeneticSearch(
/home/docs/checkouts/readthedocs.org/user_builds/mloptimizer/checkouts/master/examples/plot_mlp_neural_network.py:52: UserWarning: Some hyperparameters have very small integer ranges (< 10 distinct values): 'activation' (3 values: 0 to 2). Small ranges limit search granularity. Consider increasing the range or scale for float types.
  opt = GeneticSearch(
Starting MLPClassifier optimization...

Genetic execution:   0%|          | 0/6 [00:00<?, ?it/s, best fitness=?]
Genetic execution:  17%|█▋        | 1/6 [00:00<00:01,  2.59it/s, best fitness=0.976]
Genetic execution:  17%|█▋        | 1/6 [00:00<00:04,  1.10it/s, best fitness=0.98]
Genetic execution:  33%|███▎      | 2/6 [00:02<00:05,  1.29s/it, best fitness=0.98]
Genetic execution:  50%|█████     | 3/6 [00:04<00:04,  1.43s/it, best fitness=0.98]
Genetic execution:  67%|██████▋   | 4/6 [00:05<00:02,  1.38s/it, best fitness=0.98]
Genetic execution:  83%|████████▎ | 5/6 [00:06<00:01,  1.31s/it, best fitness=0.98]
Genetic execution: 100%|██████████| 6/6 [00:07<00:00,  1.27s/it, best fitness=0.98]
Genetic execution: 100%|██████████| 6/6 [00:09<00:00,  1.51s/it, best fitness=0.98]

GeneticSearch(cv=StratifiedKFold(n_splits=3, random_state=42, shuffle=True),
              estimator_class=<class 'sklearn.neural_network._multilayer_perceptron.MLPClassifier'>,
              generations=5,
              hyperparam_space=HyperparameterSpace(fixed_hyperparams={'max_iter': 500, 'hidden_layer_sizes': [100]}, evolvable_hyperparams={'alpha': Hyperparam('alpha', 1, 1000, 'float', 10000), 'learning_rate_init': Hyperparam('learning_rate_init', 1, 100, 'float', 10000), 'activation': Hyperparam('activation', 0, 2, 'list', ['relu', 'tanh', 'logistic'])}),
              n_elites=2, population_size=8, scoring='accuracy', seed=42,
              use_parallel=False)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

Evaluate the optimized model

best_clf = opt.best_estimator_
y_pred = best_clf.predict(X_test)
test_accuracy = accuracy_score(y_test, y_pred)
test_f1 = f1_score(y_test, y_pred, average='binary')

print(f"\nOptimization completed!")
print(f"Best parameters: {opt.best_params_}")
print(f"Test accuracy: {test_accuracy:.4f}")
print(f"Test F1: {test_f1:.4f}")

Optimization completed!
Best parameters: {'activation': 'relu', 'alpha': 0.076, 'batch_size': 'auto', 'beta_1': 0.9, 'beta_2': 0.999, 'early_stopping': False, 'epsilon': 1e-08, 'hidden_layer_sizes': [100], 'learning_rate': 'constant', 'learning_rate_init': 0.0036, 'max_fun': 15000, 'max_iter': 500, 'momentum': 0.9, 'n_iter_no_change': 10, 'nesterovs_momentum': True, 'power_t': 0.5, 'random_state': 42, 'shuffle': True, 'solver': 'adam', 'tol': 0.0001, 'validation_fraction': 0.1, 'verbose': False, 'warm_start': False}
Test accuracy: 0.9561
Test F1: 0.9645

Visualize the search space

population_df = opt.populations_
top_params = ['alpha', 'learning_rate_init', 'activation', 'fitness']
df_filtered = population_df[top_params]
g_search_space = plotly_search_space(df_filtered, top_params)
g_search_space.update_layout(
    title="MLPClassifier Hyperparameter Search Space",
    autosize=True,
    width=None,
    height=650
)
plotly.io.show(g_search_space, config={'responsive': True})

Visualize the optimization evolution

g_logbook = plotly_logbook(opt.logbook_, population_df)
g_logbook.update_layout(
    title="MLPClassifier Optimization Evolution",
    autosize=True,
    width=None,
    height=500
)
plotly.io.show(g_logbook, config={'responsive': True})

Analyze optimization performance

print("\n=== Optimization Performance ===")
print(f"Unique evaluations performed: {opt.n_trials_}")
print(f"Total individuals in population history: {len(opt.populations_)}")
print(f"Optimization time: {opt.optimization_time_:.4f} seconds")
print(f"Time per evaluation: {opt.optimization_time_ / opt.n_trials_:.4f} seconds")

=== Optimization Performance ===
Unique evaluations performed: 33
Total individuals in population history: 48
Optimization time: 9.0605 seconds
Time per evaluation: 0.2746 seconds