AutoML & Neural Architecture Search

Automated model design and hyperparameter optimization

Let algorithms find the best architecture and hyperparameters. Supports differentiable NAS (DARTS), evolutionary search, Bayesian optimization, Hyperband, and complete AutoML pipelines with feature selection, model selection, and ensemble building.

Model Selection Architecture Design Hyperparameter Tuning Feature Engineering Research Production ML

Neural Architecture Search

Automatically discover optimal neural network architectures.

DARTS

Differentiable Architecture Search with continuous relaxation of search space.

ENAS

Efficient NAS with parameter sharing across candidate architectures.

ProxylessNAS

Direct NAS on target hardware without proxy tasks.

Once-for-All

Train one network that supports diverse architectural configurations.

Hyperparameter Optimization

Find optimal hyperparameters efficiently.

Bayesian Optimization

Gaussian process-based optimization with acquisition functions.

Hyperband

Bandit-based approach with early stopping of poor configurations.

BOHB

Bayesian Optimization and HyperBand combining both approaches.

TPE

Tree-structured Parzen Estimator for efficient search.

Random Search

Surprisingly effective baseline for hyperparameter search.

AutoML Pipeline

End-to-end automated machine learning workflow.

Feature Selection

Automatic feature importance and selection methods.

Model Selection

Compare and select best model from algorithm library.

Ensemble Building

Automatically construct optimal model ensembles.

Cross-Validation

Robust model evaluation with multiple CV strategies.

AutoML with AiModelBuilder

using AiDotNet;

// AutoML with AiModelBuilder
var result = await new AiModelBuilder<float, float[], float>()
    .ConfigureAutoML(new AutoMLOptions(
        task: MLTask.Classification,
        timeLimit: TimeSpan.FromMinutes(30),
        metric: Metric.F1Score))
    .ConfigurePreprocessing()
    .BuildAsync(features, labels);

// AiModelBuilder selects the best model automatically
var prediction = result.Predict(newSample);

Start building with AutoML & Neural Architecture Search

All 4 implementations are included free under Apache 2.0.

Install AiDotNet Browse All Features