AutoML

51 public types in this namespace, organized by kind.

Models & Types (33)

Type Summary
ArchitectureDto Data transfer object for architecture JSON serialization.
Architecture<T> Represents a neural network architecture discovered through NAS.
AttentiveNAS<T> AttentiveNAS: Improving Neural Architecture Search via Attentive Sampling.
AutoMLEnsembleModel<T> A simple tabular ensemble model used as a facade-safe AutoML final model.
BayesianOptimizationAutoML<T, TInput, TOutput> Built-in AutoML strategy that uses a lightweight Bayesian-style surrogate to guide trial selection.
BigNAS<T> BigNAS: Scaling Up Neural Architecture Search with Big Single-Stage Models.
CompressionOptimizer<T> Automatically finds the best compression configuration for a model.
CompressionTrial<T> Represents a compression configuration to be evaluated.
DiffusionAutoML<T> AutoML for diffusion models with automatic hyperparameter optimization.
ENAS<T> Efficient Neural Architecture Search via Parameter Sharing.
EvolutionaryAutoML<T, TInput, TOutput> Built-in AutoML strategy that uses an evolutionary (genetic) approach to propose new trials.
FBNet<T> FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search.
GDAS<T> Gradient-based Differentiable Architecture Search with Gumbel-Softmax sampling.
HardwareConstraints<T> Hardware constraints for NAS.
HardwareCostModel<T> Models hardware costs for neural architecture search operations using FLOP-based estimation.
HardwareCost<T> Represents the hardware cost of an operation or architecture.
MapElitesAutoMLArchiveEntry Describes one immutable elite model specification retained by MAP-Elites AutoML.
MapElitesAutoML<T, TInput, TOutput> Built-in supervised AutoML strategy that searches a deterministic MAP-Elites archive.
MobileNetSearchSpace<T> Defines the MobileNet-based search space for neural architecture search.
MultiFidelityAutoML<T, TInput, TOutput> Built-in AutoML strategy that uses multi-fidelity (successive halving) and ASHA scheduling.
NeuralArchitectureSearch<T> Neural Architecture Search implementation with gradient-based (DARTS) support
OnceForAll<T> Once-for-All (OFA) Networks: Train Once, Specialize for Anything.
OperationDto Data transfer object for a single operation in the architecture.
PCDARTS<T> Partial Channel Connections for Memory-Efficient Differentiable Architecture Search.
ParameterRange Defines the range and type of a hyperparameter for AutoML search
PlatformCharacteristics Platform characteristics for hardware cost estimation.
ProxylessNAS<T> ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware.
RandomSearchAutoML<T, TInput, TOutput> AutoML implementation that uses random search over candidate model types and hyperparameters.
ResNetSearchSpace<T> Defines the ResNet-based search space for neural architecture search.
SearchConstraint Defines a constraint for AutoML search to limit the search space or enforce requirements.
SearchSpaceBase<T> Defines the search space for neural architecture search.
TransformerSearchSpace<T> Defines the Transformer-based search space for neural architecture search.
TrialResult Represents the result of a single trial during AutoML search.

Base Classes (5)

Type Summary
AutoMLModelBase<T, TInput, TOutput> Base class for AutoML models that automatically search for optimal model configurations
BuiltInSupervisedAutoMLModelBase<T, TInput, TOutput> Base class for built-in supervised AutoML strategies that operate on tabular Matrix/Vector tasks.
NasAutoMLModelBase<T> Base class for NAS-based AutoML models.
SupervisedAutoMLModelBase<T, TInput, TOutput> Base class for AutoML implementations that train and score supervised models.
UnsupervisedAutoMLBase<T> Base class for unsupervised AutoML search strategies (e.g., clustering AutoML, grid search).

Enums (5)

Type Summary
AttentiveSamplingStrategy Which candidate AttentiveNAS trains from the pool drawn under a FLOPs target (Wang et al.
ConstraintType Defines the types of constraints that can be applied to AutoML search.
DiffusionSchedulerType Represents the type of scheduler for diffusion sampling.
HardwarePlatform Supported hardware platforms.
NoisePredictorType Represents the type of noise predictor architecture.

Options & Configuration (8)

Type Summary
AttentiveNASConfig<T> Configuration for an AttentiveNAS sub-network.
BigNASConfig Configuration for a BigNAS sub-network.
CompressionOptimizerOptions Configuration options for the compression optimizer.
DartsOptimizerOptions Optimizer settings of the DARTS search in NeuralArchitectureSearch.
DiffusionTrialConfig<T> Configuration for a diffusion model trial in AutoML.
ENASOptions Hyperparameters of ENAS.
PCDARTSOptions Search settings of PCDARTS.
SubNetworkConfig Configuration for a sub-network sampled from OFA.