Interfaces

437 public types in this namespace, organized by kind.

Models & Types (91)

Type Summary
AcousticCharacteristics<T> Acoustic characteristics of a scene.
AudioEffectParameter<T> Represents an adjustable parameter for an audio effect.
AudioEventResult<T> Result of audio event detection.
AudioEvent<T> A detected audio event.
AudioFingerprint<T> Represents an audio fingerprint.
AudioVisualEvent Represents an audio-visual event with temporal boundaries.
AxisRelation How ONE output axis is sized from the input axes and the declaring type’s own configuration.
BeatTrackingResult<T> Result of beat tracking.
ChordRecognitionResult<T> Result of chord recognition.
ChordSegment<T> A segment of audio with a detected chord.
ChordStatistics<T> Statistics for a chord in the recognition result.
Contradiction Represents a detected contradiction between reasoning steps.
CritiqueResult<T> Result of critiquing a reasoning step or chain.
DataLoaderCheckpoint Serializable checkpoint for a stateful data loader.
DatasetMetadata Metadata about a dataset.
DiarizationResult<T> Result of speaker diarization.
DirectionEstimate<T> A direction estimate without full source information.
DownbeatResult<T> Result of downbeat detection.
EmotionResult<T> Represents the result of emotion recognition.
EpisodeResult<T> Result of running a single RL episode.
EventStatistics<T> Statistics for an event type.
FewShotExample Represents a single few-shot example with input and output.
FingerprintMatch Represents a match between two fingerprints.
FrequencyRange<T> A frequency range.
GeneticParameters Parameters for configuring a genetic algorithm.
GenreClassificationResult<T> Result of genre classification.
GenreFeatures<T> Features extracted for genre classification (generic version).
GenrePrediction<T> A single genre prediction with confidence.
GenreSegment<T> A segment with genre information.
GenreTrackingResult<T> Result of tracking genre over time.
GpuOptimizerState Holds the optimizer state buffers for GPU-resident training.
IncrementalState<T> Represents the internal state of a time series transformer for incremental processing.
KeyDetectionResult<T> Result of key detection.
KeyHypothesis<T> A key hypothesis with confidence score.
KeySegment<T> A segment with a detected key.
KeyTrackingResult<T> Result of key tracking over time.
LanguageResult<T> Represents the result of language identification.
LanguageSegment<T> Represents a time-segmented language detection result.
LayerInfo<T> Metadata about a single layer within a layered model, including its position in the flat parameter vector and computational characteristics.
LocalObjectiveContext<T> One hidden layer’s view for BuildLocalObjective.
LocalObjectiveStep<T> State shared by every local objective of one training step.
LocalizationResult<T> Result of sound localization.
MedicalSegmentationResult<T> Result of medical image segmentation.
Mesh3D<T> Represents a 3D mesh with vertices, faces, and optional textures.
ModelDownloadProgress Progress information for model downloads.
ModulationPoint<T> A point where the key changes (modulation).
MultimodalInput<T> Represents an input item for unified multimodal models.
MultimodalOutput<T> Represents an output from unified multimodal models.
OpenVocabSegmentationResult<T> Result of open-vocabulary segmentation.
OptimizationHistoryEntry<T> Represents a single entry in the optimization history.
OverlapRegion<T> Represents a region where speakers overlap.
PanopticSegment<T> A single segment in a panoptic segmentation result.
PanopticSegmentationResult<T> Result of panoptic segmentation containing both semantic and instance information.
PipelineOperation Represents a single operation in the pipeline schedule.
PitchFrame<T> Represents a single pitch detection frame.
PromptedSegmentationResult<T> Result from a prompted segmentation operation containing one or more mask proposals.
ReasoningContext Context information for critiquing reasoning steps.
ReferringSegmentationResult<T> Result of referring segmentation.
ResamplingStatistics<T> Contains statistics about a resampling operation.
SceneClassificationResult<T> Result of scene classification.
SceneFeatures<T> Features extracted for scene classification (generic version).
ScenePrediction<T> A single scene prediction with confidence.
SceneSegment<T> A segment with scene information.
SceneTrackingResult<T> Result of tracking scene changes over time.
SceneTransition<T> A detected scene transition.
SegmentedStructure Metadata about a single segmented medical structure.
SeparationQuality<T> Quality metrics for source separation.
SoundSource<T> A detected sound source with position.
SoundTrackingResult<T> Result of sound source tracking over time.
SourceSeparationResult<T> Result of source separation.
SourceTrajectory<T> A tracked trajectory of a sound source.
SpeakerProfile<T> Represents an enrolled speaker profile.
SpeakerSegment<T> Represents a speaker segment in diarization output.
SpeakerStatistics<T> Statistics for a speaker in diarization output.
SpeakerTurn<T> Represents a speaker turn in diarization output (legacy API compatibility).
SpeakerVerificationResult<T> Result of a speaker verification attempt.
SubModel<T> Represents a contiguous sub-model extracted from a larger ILayeredModel.
TempoHypothesis<T> A tempo hypothesis with confidence score.
TimeSignature Represents a musical time signature.
TimedEmotionResult<T> Represents a timed emotion prediction.
TrainingProgress<T> A snapshot of training progress for a single completed epoch, handed to every registered ITrainingCallback so callers can plot live loss curves, stream metrics to a dashboard, or decide whether training should keep going.
TrajectoryPoint<T> A single point in a source trajectory.
TranscribedNote<T> Represents a single transcribed musical note.
TranscriptionResult<T> Represents the result of a transcription operation.
TranscriptionSegment<T> Represents a segment of transcribed text with timing information.
TransformerState<T> Represents the serializable state of a fitted time series transformer.
VerificationResult<T> Represents the result of external tool verification.
VideoSegmentationResult<T> Result of video segmentation for a single frame.
VoiceInfo<T> Information about an available TTS voice.
WeightLoadResult Result of weight loading operation.
WeightLoadValidation Result of weight validation.

Interfaces (308)

Type Summary
I2DInterpolation<T> Defines an interface for two-dimensional interpolation algorithms.
I3DDiffusionModel<T> Interface for 3D diffusion models that generate 3D content like point clouds, meshes, and scenes.
IActivationFunction<T> Defines an interface for activation functions used in neural networks and other machine learning algorithms.
IActiveLearningStrategy<T> Defines a strategy for active learning that selects the most informative samples for labeling from a pool of unlabeled data.
IAdaptedMetaModel<T> Extended interface for meta-learning adapted models that carry task-specific adaptation state beyond backbone parameters.
IAdversarialAttack<T, TInput, TOutput> Defines the contract for adversarial attack algorithms that generate adversarial examples.
IAdversarialDefense<T, TInput, TOutput> Defines the contract for adversarial defense mechanisms that protect models against attacks.
IAdversarialInputPreprocessor<T> An input transformation applied before inference to remove adversarial perturbations.
IAggregationStrategy<TModel> Defines strategies for aggregating model updates from multiple clients in federated learning.
IAiModelBuilder<T, TInput, TOutput> Defines a builder pattern interface for creating and configuring predictive models.
IAlignmentMethod<T> Defines the contract for AI alignment methods that ensure models behave according to human values and intentions.
IAnomalyDetector<T> Defines methods for algorithmic anomaly/outlier detection using a fit-predict pattern.
IAssociativeMemory<T> Interface for Associative Memory modules used in nested learning.
IAsyncTreeBasedModel<T> Defines an interface for asynchronous tree-based machine learning models.
IAttentionBlockDecorator<T> A layer that WRAPS an attention block, adding something to its output without replacing how the block itself is invoked.
IAudioCodec<T> Defines the contract for neural audio codecs that compress and decompress audio.
IAudioDiffusionModel<T> Interface for audio diffusion models that generate sound and music.
IAudioEffect<T> Defines the contract for audio effects processors.
IAudioEnhancer<T> Defines the contract for audio enhancement models that improve audio quality.
IAudioEventDetector<T> Interface for audio event detection models that identify specific sounds/events in audio.
IAudioFeatureExtractor<T> Defines the contract for audio feature extraction algorithms.
IAudioFingerprinter<T> Interface for audio fingerprinting algorithms.
IAudioFoundationModel<T> Defines the contract for self-supervised audio foundation models.
IAudioGenerator<T> Interface for audio generation models that create audio from text descriptions or other conditions.
IAudioLanguageModel<T> Interface for multimodal audio-language models that understand and reason about audio.
IAudioSafetyModule<T> Interface for safety modules that operate on audio content.
IAudioVisualCorrespondenceModel<T> Defines the contract for audio-visual correspondence learning models.
IAudioVisualEventLocalizationModel<T> Defines the contract for audio-visual event localization models.
IAutoMLModel<T, TInput, TOutput> Defines the contract for AutoML models that automatically search for optimal model configurations.
IAutoregressiveMultimodalModel<T> Defines the contract for autoregressive multimodal generation models that can generate tokens from any modality in an interleaved fashion.
IAuxiliaryLossLayer<T> Interface for neural network layers that report auxiliary losses in addition to the primary task loss.
IBatchAwareShapeContract Opt-in for a contract whose answer depends on whether the incoming rank INCLUDES a batch axis.
IBatchIterable<TBatch> Defines capability to iterate through data in batches.
IBatchSampler Extended interface for samplers that support batch-level sampling.
IBeatTracker<T> Interface for beat tracking models that detect tempo and beat positions in audio.
IBenchmark<T> Defines the contract for reasoning benchmarks that evaluate model performance.
IBiasDetector<T> Defines an interface for detecting bias in machine learning model predictions.
IBlip2Model<T> Defines the contract for BLIP-2 (Bootstrapped Language-Image Pre-training 2) models.
IBlipModel<T> Defines the contract for BLIP (Bootstrapped Language-Image Pre-training) models.
ICausalModel<T> Interface for causal inference models (meta-learners).
ICertifiedDefense<T, TInput, TOutput> Defines the contract for certified defense mechanisms that provide provable robustness guarantees.
IChatClientDecorator<T> A chat client that wraps another one and can hand out the client it wraps.
ICheckpointManager<T, TInput, TOutput> Defines the contract for checkpoint management systems that save and restore training state.
ICheckpointableModel Defines the contract for models that support saving and loading their internal state (checkpointing).
IChordRecognizer<T> Interface for chord recognition models that identify musical chords in audio.
IChunkingStrategy Defines the contract for text chunking strategies that split documents into smaller segments.
IClassifier<T> Defines the common interface for all classification algorithms in the AiDotNet library.
IClientModel<TData, TUpdate> Defines the functionality for a client-side model in federated learning.
IClientSelectionStrategy Selects which clients participate in a federated learning round.
ICloneable<T> Interface for objects that can be cloned or copied.
ICompositeLoss<T> A model whose training objective is a WEIGHTED SUM over several named outputs, declared rather than hand-assembled at each call site.
ICompressionMetadata<T> Defines the contract for compression metadata that stores information needed to decompress model weights.
IConditioningModule<T> Interface for conditioning modules that encode various inputs into embeddings for diffusion models.
IContextCompressor<T> Defines the contract for compressing context documents to reduce token usage while preserving relevance.
IContextFlow<T> Interface for Context Flow mechanism - maintains distinct information pathways and update rates for each nested optimization level.
IContinualLearningStrategy<T> Defines a strategy for continual learning that helps neural networks learn multiple tasks sequentially without forgetting previously learned knowledge.
IContrastiveLoss<T> Interface for contrastive and self-supervised loss functions that operate on pairs of embeddings/representations rather than predictions vs ground truth labels.
ICountable Defines capability to report dataset size and iteration progress.
ICreditAssignmentContext<T> The information an ICreditRule receives for a single training step: the ordered trainable layers (with output shapes and current weights) and the network output error signal.
ICreditLayer<T> A single trainable layer as seen by an ICreditRule: its output shape, its weight matrix (when it has a single one), and the teaching-signal slot the rule fills.
ICreditRule<T> A pluggable credit-assignment (a.k.a.
ICrossValidator<T, TInput, TOutput> Defines the contract for cross-validation implementations in machine learning models.
IDallE3Model<T> Defines the contract for DALL-E 3-style text-to-image generation models.
IDataLoader<T> Base interface for all data loaders providing common data loading capabilities.
IDataLoader<TInput, TOutput> Data loader parameterized by both input and output tensor types — the type-theoretic fit for domains where the batch shape is richer than the row-scalar (X, y) supervised pair (image-space photometric NeRF training, video / audio / sequence…
IDataSampler Defines the contract for sampling indices from a dataset during batch iteration.
IDataTransformer<T, TInput, TOutput> Defines a data transformer that can fit to data and transform it.
IDataVersionControl<T> Defines the contract for data version control systems that track dataset changes over time.
IDatasetFactory<T, TInput, TOutput> Factory for creating datasets.
IDataset<T, TInput, TOutput> Interface for datasets used in active learning scenarios.
IDecisionFunctionClassifier<T> Interface for classifiers that compute a decision function for predictions.
IDerivativeFreeFunctionOptimizer<T> Interface for optimizing a scalar-valued function over a vector of parameters when no gradient is available.
IDetectionBackbone<T> Marker interface for detection backbones.
IDetectionTrainingModel<T> A model that implements its detection task’s assignment and classification/box loss.
IDiagnosticsProvider Interface for components that provide diagnostic information for monitoring and debugging.
IDiffusionModel<T> Interface for diffusion-based generative models.
IDisposableGaussianProcess<T> A Gaussian Process whose implementation owns native or unmanaged resources (e.g., cached Cholesky factors, GPU memory) and must be released deterministically.
IDistillationStrategy<T> Defines a strategy for computing knowledge distillation loss between student and teacher models.
IDocumentStore<T> Defines the contract for document stores that index and retrieve vectorized documents.
IDomainReasoner<T> Interface for domain-specific reasoning models that solve problems using LLM-based reasoning strategies (chain-of-thought, tree search, consensus).
IDownloadable Defines capability to automatically download and cache datasets.
IDriftDetector<T> Defines the interface for concept drift detection.
IEmbeddingClient Turns text into embedding vectors, reporting transport failure as a value rather than an exception.
IEmbeddingModel<T> Defines the contract for embedding models that convert text into vector representations.
IEmotionRecognizer<T> Defines the contract for speech emotion recognition models.
IEnvironment<T> Represents a reinforcement learning environment that an agent interacts with.
IEpisode<T, TInput, TOutput> Represents a single episode in meta-learning, wrapping an IMetaLearningTask with additional metadata such as domain, difficulty, and timing information.
IEpisodicDataLoader<T, TInput, TOutput> Interface for data loaders that provide episodic tasks for meta-learning.
IEpisodicDataset<T, TInput, TOutput> Interface for episodic datasets used in meta-learning.
IEvolvable<TGene, T> Represents an individual that can evolve through genetic operations.
IExogenousForecastModel<T> Marks a time-series model that forecasts from future exogenous regressors (e.g.
IExperiment Represents a machine learning experiment that groups related training runs.
IExperimentRun<T> Represents a single training run within an experiment.
IExperimentTracker<T> Defines the contract for experiment tracking systems that log machine learning experiments.
IExtendedDataset<T, TInput, TOutput> Extended dataset interface with additional metadata and features.
IFairnessEvaluator<T> Defines an interface for evaluating fairness in machine learning models.
IFeatureAware Interface for models that can provide information about their feature usage.
IFeatureImportance<T> Interface for models that can provide feature importance scores.
IFeatureMapProvider<T> Mixin interface for neural networks that produce multi-scale feature pyramids — typically detection / segmentation backbones (ResNet, CSPDarknet, EfficientNet, SwinTransformer) whose outputs feed FPN, PAN, anchor generators, or DETR-style t…
IFeatureSelector<T> Defines a fitted transformer that keeps a subset of its input’s columns: a feature selector.
IFederatedClientDataLoader<T, TInput, TOutput> Represents a data loader that can provide per-client datasets for federated learning.
IFederatedHeterogeneityCorrection<T> Applies a heterogeneity correction transform to client updates in federated learning.
IFederatedServerOptimizer<T> Applies a server-side optimization step in federated learning (FedOpt family).
IFederatedTrainer<TModel, TData, TMetadata> Defines the core functionality for federated learning trainers that coordinate distributed training across multiple clients.
IFeedbackLearningRule<T> Optional extension of ICreditRule for rules whose feedback state is learned across training steps (e.g.
IFewShotExampleSelector<T> Defines the contract for selecting few-shot examples to include in prompts.
IFineTuning<T, TInput, TOutput> Defines the contract for fine-tuning methods that adapt pre-trained models to specific tasks or preferences.
IFitDetector<T, TInput, TOutput> Defines an interface for detecting how well a machine learning model fits the data.
IFitnessCalculator<T, TInput, TOutput> Defines an interface for calculating how well a machine learning model performs.
IFixedCountFeatureSelector<T> Defines a feature selector configured to keep a fixed number of columns.
IFlamingoModel<T> Defines the contract for Flamingo-style models with in-context visual learning capabilities.
IFullModel<T, TInput, TOutput> Represents a complete machine learning model that combines prediction capabilities with serialization and checkpointing support.
IFunctionOptimizer<T> Interface for optimizing a scalar-valued function over a vector of parameters.
IFusedReadoutLossLayer<T> A readout layer that can compute loss(Forward(input), target) as one fused, differentiable operation without materialising its output on the autodiff tape - for example a language-model head fused with cross-entropy (`IEngine.FusedLinearC…
IGaussianProcessClassifier<T> Defines an interface for Gaussian Process classification, a probabilistic approach to classification that provides uncertainty estimates along with class predictions.
IGaussianProcess<T> Defines an interface for Gaussian Process regression, a powerful probabilistic machine learning technique.
IGenerator<T> Defines the contract for text generation models used in retrieval-augmented generation.
IGeneticAlgorithm<T, TInput, TOutput, TIndividual, TGene> Represents a machine learning model that uses genetic algorithms or evolutionary computation while maintaining the core capabilities of a full model.
IGenreClassifier<T> Interface for music genre classification models.
IGlobalExplainer<T, TExplanation> Interface for explainers that provide global (model-wide) explanations.
IGpt4VisionModel<T> Defines the contract for GPT-4V-style models that combine vision understanding with large language model capabilities.
IGpuOptimizerConfig Configuration for GPU-resident optimizer updates.
IGradientBasedOptimizer<T, TInput, TOutput> Defines an interface for optimization algorithms that use gradients to find the best parameters for a model.
IGradientCache<T> Defines an interface for storing and retrieving pre-computed gradients to improve performance in machine learning models.
IGradientComputable<T, TInput, TOutput> Base interface for models that can compute gradients explicitly without updating parameters.
IGradientModel<T> Represents a gradient for optimization algorithms.
IGraphConvolutionLayer<T> Defines the contract for graph convolutional layers that process graph-structured data.
IGraphDataLoader<T> Interface for data loaders that provide graph-structured data for graph neural networks.
IGraphInferenceModel<T> Marker interface for models that support graph-aware inference: bind an adjacency matrix, forward node features, get per-node predictions.
IGraphStore<T> Defines the contract for graph storage backends that manage nodes and edges.
IHomomorphicEncryptionProvider<T> Provides homomorphic encryption operations for federated learning aggregation.
IHyperparameterOptimizer<T, TInput, TOutput> Defines the contract for hyperparameter optimization algorithms.
IImageBindModel<T> Defines the contract for ImageBind models that bind multiple modalities (6+) into a shared embedding space.
IImageSafetyModule<T> Interface for safety modules that operate on image content.
IInputGradientComputable<T> Interface for models that support computing gradients with respect to input data.
IInputOutputDataLoader<T, TInput, TOutput> Interface for data loaders that provide standard input-output (X, Y) data for supervised learning.
IInstanceSegmentation<T> Interface for instance segmentation models that detect and mask individual object instances.
IIntermediateActivationStrategy<T> Defines methods for distillation strategies that utilize intermediate layer activations.
IInterpolation<T> Defines an interface for interpolation algorithms that estimate values between known data points.
IInterpretableModel<T> Interface for models that support interpretability features.
IKernelFunction<T> Defines an interface for kernel functions that measure similarity between data points in machine learning algorithms.
IKeyDetector<T> Interface for musical key detection models that identify the key and mode of music.
IKnowledgeDistillationTrainer<T, TInput, TOutput> Defines the contract for knowledge distillation trainers that train student models using knowledge transferred from teacher models.
ILLaVAModel<T> Defines the contract for LLaVA (Large Language and Vision Assistant) models.
ILanguageIdentifier<T> Defines the contract for spoken language identification from audio.
ILanguageModel<T> Marker interface for models with a language-model-shaped inference contract: token-tensor input [B, T] → next-token-logits output [B, T, V].
ILatentDiffusionModel<T> Interface for latent diffusion models that operate in a compressed latent space.
ILayer<T> Defines the contract for neural network layers in the AiDotNet framework.
ILayeredModel<T> Provides layer-level access to a neural network’s architecture and parameters.
ILinearRegression<T> Defines an interface for linear regression in machine learning, which predict outputs as a weighted sum of inputs plus an optional constant.
ILinkFunction<T> Interface for link functions used in Generalized Linear Models (GLMs).
ILoRAAdapter<T> Interface for LoRA (Low-Rank Adaptation) adapters that wrap existing layers with parameter-efficient adaptations.
ILoRAConfiguration<T> Interface for configuring how LoRA (Low-Rank Adaptation) should be applied to neural network layers.
ILocalExplainer<T, TExplanation> Interface for explainers that provide local (per-instance) explanations.
ILocalObjectiveCreditRule<T> A credit rule that trains every hidden layer from a local objective — a scalar loss defined on that layer’s own output — instead of a teaching signal routed back from the network output.
ILossFunction<T> Interface for loss functions used in neural networks.
IMapElitesAutoMLModel<T, TInput, TOutput> Adds immutable quality-diversity archive inspection to the standard AutoML contract.
IMaskedActionEnvironment<T> An environment that can say which discrete actions are legal in the current state.
IMatrixDecomposition<T> Represents a matrix decomposition that can be used to solve linear systems and invert matrices.
IMedicalSegmentation<T> Interface for medical image segmentation models that handle 2D slices and 3D volumetric data.
IMetaDataset<T, TInput, TOutput> Represents a high-level meta-dataset that can generate episodes for meta-learning.
IMetaLearnerOptions<T> Configuration options interface for meta-learning algorithms.
IMetaLearner<T, TInput, TOutput> Unified interface for meta-learning algorithms that train models to quickly adapt to new tasks.
IMetaLearningTask<T, TInput, TOutput> Represents a single meta-learning task for few-shot learning.
IModelCache<T, TInput, TOutput> Defines a caching mechanism for storing and retrieving optimization step data during model training.
IModelCompressionStrategy<T> Defines an interface for model compression strategies used to reduce model size while preserving accuracy.
IModelCompression<T, TMetadata> Defines a type-safe interface for model compression used to reduce model size while preserving accuracy.
IModelExplainer<T> Interface for model-agnostic explainers that can explain any predictive model’s decisions.
IModelRegistry<T, TInput, TOutput> Defines the contract for model registry systems that manage trained model storage and versioning.
IModelSerializer Defines methods for converting machine learning models to and from binary data for storage or transmission.
IModelShape Provides shape metadata for a machine learning model, describing its expected input and output dimensions.
IModel<TInput, TOutput, TMetadata> Defines the core functionality for machine learning models that can be trained on data and make predictions.
IMultiLabelClassifier<T> Interface for multi-label classification models.
IMultiObjectiveIndividual<T> Interface for individuals supporting multi-objective optimization.
IMultiOutputShapeContract Opt-in for a contract describing a layer that returns MORE THAN ONE tensor.
IMultiPortShapeContract Opt-in for a contract whose output depends on MORE THAN ONE input port.
IMultimodalEmbedding<T> Interface for multimodal embedding models that can encode multiple modalities (text, images, audio) into a shared embedding space.
IMultivariateForecastModel<T> Marks a multivariate time-series model that forecasts several interrelated series at once (e.g.
IMusicSourceSeparator<T> Interface for music source separation models that isolate individual instruments/vocals from a mix.
IMusicTranscriber<T> Defines the contract for automatic music transcription (audio to notes).
INeuralNetworkModel<T> Defines the contract for neural network models with advanced architectural introspection capabilities.
INeuralNetwork<T> Defines the core functionality for neural network models in the AiDotNet library.
INoisePredictor<T> Interface for noise prediction networks used in diffusion models.
INoiseSchedulerInputScaling<T> Optional sampler capability for schedules whose latent coordinate system differs from the denoiser’s model-input coordinate system, such as sigma-space Euler sampling.
INoiseScheduler<T> Interface for diffusion model noise schedulers that control the noise schedule during inference.
INonLinearRegression<T> Defines the functionality for non-linear regression models in the AiDotNet library.
IOnlineClassifier<T> Interface for online (incremental) classification models.
IOnlineLearningModel<T> Defines the interface for online (incremental) learning models.
IOnnxModelDownloader Defines the contract for downloading ONNX models from remote sources.
IOnnxModelMetadata Metadata about an ONNX model including inputs, outputs, and opset version.
IOnnxModel<T> Defines the contract for ONNX model wrappers that provide cross-platform model inference.
IOnnxTensorInfo Information about an ONNX tensor (input or output).
IOpenVocabSegmentation<T> Interface for open-vocabulary segmentation models that segment objects from text descriptions without being limited to a fixed set of classes.
IOptimizer<T, TInput, TOutput> Defines the contract for optimization algorithms used in machine learning models.
IOrdinalClassifier<T> Interface for ordinal classification (ordinal regression) models.
IOutlierDetector<T> Defines methods for algorithmic outlier/anomaly detection using a fit-predict pattern.
IOutlierRemoval<T, TInput, TOutput> Defines methods for detecting and removing outliers from datasets.
IOutputDerivative<T> Interface for activation functions that can compute their derivative given the post-activation output value rather than the pre-activation input.
IPanopticSegmentation<T> Interface for panoptic segmentation models that unify semantic and instance segmentation.
IPaperOptimizerVariant Implemented by models whose paper specifies different hyperparameters per size or configuration variant, so the right PaperOptimizerAttribute can be selected.
IParameterSource<T> Anything that owns trainable parameters and can hand them out and take them back as a flat vector: a model, a sub-model, a layer, or an internal component such as an encoder or projector.
IParameterizable<T, TInput, TOutput> Interface for models that have optimizable parameters.
IPipelineDecomposableModel<T, TInput, TOutput> Interface for models that support decomposing the backward pass into separate activation gradient and weight gradient computations.
IPipelinePartitionStrategy<T> Defines a strategy for partitioning model parameters across pipeline stages.
IPipelineSchedule<T> Defines a scheduling strategy for pipeline parallel training.
IPipelineStep<T, TInput, TOutput> Represents a step in a data processing pipeline
IPitchDetector<T> Defines the contract for pitch (fundamental frequency) detection.
IPostprocessor<T, TInput, TOutput> Defines a postprocessor that transforms model outputs into final results.
IPredictiveModel<T, TInput, TOutput> Defines the core functionality of a trained predictive model that can make predictions on new data.
IPreferredDataSetFitnessCalculator<T, TInput, TOutput> Optional extension interface for fitness calculators that have a preferred dataset type.
IPrivacyAccountant Tracks cumulative privacy loss across federated learning rounds.
IPrivacyMechanism<TModel> Defines privacy-preserving mechanisms for federated learning to protect client data.
IProbabilisticClassifier<T> Defines the interface for classifiers that can output probability estimates for each class.
IProgramArtifactStore Stores and retrieves the evidence an evaluation produced, keyed by the genome that produced it.
IProgramDescriptor Computes one named, finite behaviour descriptor from a candidate program’s text alone.
IProgramFitnessEvaluator Scores one candidate program and reports the result as an evolution task result.
IProgramNoveltyJudge Decides whether a candidate program differs meaningfully from the incumbent it most resembles.
IPromptAnalyzer Defines the contract for analyzing prompts before sending them to language models.
IPromptCompressor Defines the contract for compressing prompts to reduce token counts and API costs.
IPromptOptimizer<T> Defines the contract for optimizing prompts to improve language model performance.
IPromptTemplate Defines the contract for prompt templates used in language model interactions.
IPromptableSegmentation<T> Interface for interactive, promptable segmentation models like SAM that accept user prompts (points, boxes, masks, text) to segment specific objects.
IProposalProvenanceSink Receives one provenance record per language-model request made while proposing a candidate program.
IPruningMask<T> Represents a binary mask for pruning weights in a neural network layer.
IPruningStrategy<T> Interface for pruning strategies that remove unimportant weights to create sparsity.
IQkvTransform<T> Transforms an attention block’s projected queries, keys and values BEFORE they are split into heads and scored.
IQueryProcessor Defines the contract for processing and transforming user queries before retrieval.
IRAGMetric<T> Defines the contract for RAG evaluation metrics.
IRLAgent<T> Marker interface for reinforcement learning agents that integrate with AiModelBuilder.
IRLDataLoader<T> Interface for data loaders that provide experience data for reinforcement learning.
IRadialBasisFunction<T> Defines a radial basis function (RBF) that measures similarity based on distance.
IReasoningModelDiagnosticSink Receives the record of every adjustment a reasoning-model profile makes to an outgoing request.
IReasoningStrategy<T> Defines the contract for reasoning strategies that solve problems through structured thinking.
IRebasableProgramDescriptor A program descriptor that measures against other programs and can be pointed at a new set of them.
IRecognitionTrainingModel<T> A text recogniser that trains on transcriptions, using its paper’s objective (CTC for CRNN, teacher-forced cross-entropy for TrOCR).
IReferringSegmentation<T> Interface for referring segmentation models that segment objects based on natural language descriptions, including complex reasoning about spatial relationships and attributes.
IRegression<T> Defines the common interface for all regression algorithms in the AiDotNet library.
IRegularization<T, TInput, TOutput> Defines methods for applying regularization techniques to machine learning models.
IReranker<T> Defines the contract for reranking retrieved documents to improve relevance ordering.
IResamplingStrategy<T> Defines the interface for resampling strategies used to handle imbalanced datasets.
IResettable Defines capability to reset iteration state back to the beginning.
IRetriever<T> Defines the contract for retrieving relevant documents based on a query.
ISafetyFilter<T> Defines the contract for safety filters that detect and prevent harmful or inappropriate model inputs and outputs.
ISafetyModule<T> Base interface for all safety modules in the composable safety pipeline.
ISceneClassifier<T> Interface for acoustic scene classification models that identify the environment/context of audio.
ISecondOrderGradientComputable<T, TInput, TOutput> Extended gradient computation interface for MAML meta-learning algorithms.
ISegmentationModel<T> Base interface for all image segmentation models that classify pixels into categories.
ISelfSupervisedLoss<T> Interface for self-supervised loss functions used in meta-learning.
ISelfSupervisedModel Marker interface for models that train themselves from unlabeled data — the model owns its training objective rather than fitting a supervised target.
ISemanticSegmentation<T> Interface for semantic segmentation models that assign a class label to every pixel.
ISemiSupervisedClassifier<T> Defines the interface for semi-supervised classification algorithms that can learn from both labeled and unlabeled data.
ISequenceLossFunction<T> Interface for sequence loss functions that operate on variable-length sequences.
IShapeContract Declares, symbolically, how a layer or model sizes each of its output axes.
ISharedParameterConsumerLayer<T> Optional contract for layers that reference trainable parameter tensors owned by another layer (such as a tied output head reading an embedding layer’s table, or a shared-substrate child block referencing a root block’s weights) without re-…
IShuffleable Defines capability to shuffle data for randomized iteration.
ISoundLocalizer<T> Interface for sound localization models that estimate the spatial position of sound sources.
ISpeakerDiarizer<T> Interface for speaker diarization models that segment audio by speaker (“who spoke when”).
ISpeakerEmbeddingExtractor<T> Interface for speaker embedding extraction models (d-vector/x-vector extraction).
ISpeakerVerifier<T> Interface for speaker verification models that determine if audio matches a claimed identity.
ISpeechRecognizer<T> Interface for speech recognition models that transcribe audio to text (ASR - Automatic Speech Recognition).
IStatefulDataLoader<T> Extends IDataLoader with checkpoint/resume support for fault-tolerant training.
IStratifiedSampler Interface for samplers that use class labels for stratification.
IStreamingDataLoader<T, TInput, TOutput> Interface for streaming data loaders that process data on-demand without loading all data into memory.
IStreamingEventDetectionSession<T> Interface for streaming event detection sessions.
IStreamingGenerator<T> A generator that can stream its output incrementally (token/chunk deltas) as it is produced, in addition to the buffered Generate.
IStreamingSynthesisSession<T> Interface for streaming TTS synthesis sessions.
IStreamingTranscriptionSession<T> Interface for streaming transcription sessions.
ISupervisedFeatureSelector<T> Defines a feature selector that needs the target values to decide which columns to keep.
ISupportsLossFunction<T> Implemented by models whose loss function can be replaced by the caller after construction.
ISupportsRepresentation<T, TInput> An opt-in capability for models that can expose a dense internal representation (embedding) of an input — typically the penultimate-layer activations.
ISurvivalModel<T> Interface for survival analysis models.
ISyntheticTabularGenerator<T> Defines the contract for synthetic tabular data generators that learn the distribution of real tabular data and can produce new, realistic synthetic rows.
ITaskSampler<T, TInput, TOutput> Controls the strategy for sampling meta-learning tasks (episodes) from a dataset.
ITeacherModel<TInput, TOutput> Represents a trained teacher model for knowledge distillation.
ITextDetectionTrainingModel<T> A text detector that trains on text polygons, using its paper’s target assignment and loss (DBNet’s shrunk-kernel and border maps, CRAFT’s character and affinity heatmaps, EAST’s score and geometry maps).
ITextGenerator A minimal, numeric-type-agnostic text generator: turns a prompt into generated text.
ITextSafetyModule<T> Interface for safety modules that operate on text content.
ITextToSpeech<T> Interface for text-to-speech (TTS) models that synthesize spoken audio from text.
ITextVectorizer<T> Defines a text vectorizer that converts text documents to numeric feature matrices.
ITimeSeriesClassifier<T> Interface for time series classification models.
ITimeSeriesDecomposition<T> Defines methods and properties for decomposing time series data into its component parts.
ITimeSeriesFeatureExtractor<T> Defines a specialized data transformer for extracting features from time series data.
ITimeSeriesModel<T> Defines the core functionality for time series prediction models.
ITokenEmbeddingModel<T> Defines the contract for a model that produces per-token (contextual) embeddings for a text.
ITokenEmbedding<T> Exposes token embedding lookup for models that maintain a token embedding table.
ITrainableLayer<T> Defines a neural network layer with trainable parameters that can be used with tape-based automatic differentiation (autodiff).
ITrainer<T> Interface for training machine learning models from configuration-driven recipes.
ITrainingCallback<T> A user-registerable hook that observes supervised training epoch-by-epoch and can request that training stop early.
ITrainingEpochReporter<T> Implemented by models that run their own epoch loop internally and can report each epoch back to the caller.
ITrainingMonitor<T> Defines the contract for training monitoring systems that track and visualize model training progress.
ITreeBasedClassifier<T> Interface for tree-based classification algorithms.
ITreeBasedRegression<T> Defines the core functionality for tree-based machine learning models.
IUnifiedMultimodalModel<T> Defines the contract for unified multimodal models that handle multiple modalities in a single architecture, similar to GPT-4o, Gemini, or Meta’s CM3Leon.
IUpstreamBiasRedundancy Optional capability for a layer that can make an immediately preceding additive bias redundant.
IVAEModel<T> Interface for Variational Autoencoder (VAE) models used in latent diffusion.
IVectorActivationFunction<T> Defines activation functions that operate on vectors and tensors in neural networks.
IVersionedProgramDescriptor A program descriptor that publishes a hash of its own configuration, not only its name and type.
IVideoCLIPModel<T> Defines the contract for VideoCLIP-style models that align video and text in a shared embedding space.
IVideoDiffusionModel<T> Interface for video diffusion models that generate temporal sequences.
IVideoSafetyModule<T> Interface for safety modules that operate on video content.
IVideoSegmentation<T> Interface for video segmentation models that track and segment objects across video frames.
IVideoSuperResolution<T> The video super-resolution contract: upscale a low-resolution frame sequence.
IVideoVAEModel<T> A VAE that encodes video, [batch, channels, frames, height, width], compressing time as well as space.
IVoiceActivityDetector<T> Defines the contract for Voice Activity Detection (VAD) models.
IWaveletFunction<T> Defines the functionality for wavelet transforms used in signal processing and data analysis.
IWeightLoadable<T> Defines the contract for models that support loading weights by name.
IWeightedSampler<T> Interface for samplers that use sample weights.
IWindowFunction<T> Defines functionality for creating window functions used in signal processing and data analysis.

Enums (22)

Type Summary
ArrayType Types of microphone arrays.
ConditioningType Types of conditioning supported by diffusion models.
DallE3ImageSize Represents the available image sizes for DALL-E 3 generation.
DallE3Quality Represents the quality settings for DALL-E 3 generation.
DallE3Style Represents the style settings for DALL-E 3 generation.
DriftStatus Represents the status of drift detection.
FineTuningCategory Categories of fine-tuning methods.
FineTuningMethodType Specific fine-tuning method types.
Form The forms a relation can take.
FrameInterpolationMethod Methods for interpolating between video frames.
GpuOptimizerType Enumerates the types of GPU-optimized optimizers available.
LayerCategory Classification of neural network layer types for automated per-layer decisions.
LogLevel Log levels for training messages.
ModalityType Represents the different modality types supported by ImageBind.
ModelStage Represents the stages a model can be in during its lifecycle.
MusicalMode Musical mode (major or minor).
PipelineOperationType Types of pipeline operations.
RewardModelType Types of reward models.
SparsityPattern Types of sparsity patterns.
SurfaceReconstructionMethod Methods for reconstructing surfaces from point clouds.
TempoRelation Relationship of a tempo hypothesis to the primary tempo.
VoiceGender Gender classification for TTS voices.

Structs (2)

Type Summary
OutputAxisContract One output axis, and the relation that sizes it.
OutputSpec<T> One declared output of a multi-output model: what it is called, what shape it has, how it is scored, and how heavily it counts.

Options & Configuration (12)

Type Summary
AdagradGpuConfig Configuration for Adagrad optimizer on GPU.
AdamGpuConfig Configuration for Adam optimizer on GPU.
AdamWGpuConfig Configuration for AdamW optimizer on GPU.
AudioFeatureOptions Options for audio feature extraction.
AudioGenerationOptions<T> Advanced options for audio generation.
LambGpuConfig Configuration for LAMB (Layer-wise Adaptive Moments) optimizer on GPU.
LarsGpuConfig Configuration for LARS (Layer-wise Adaptive Rate Scaling) optimizer on GPU.
MicrophoneArrayConfig<T> Configuration for microphone array geometry.
NagGpuConfig Configuration for Nesterov Accelerated Gradient (NAG) optimizer on GPU.
PruningConfig Configuration for pruning operations.
RmsPropGpuConfig Configuration for RMSprop optimizer on GPU.
SgdGpuConfig Configuration for SGD (Stochastic Gradient Descent) optimizer on GPU.

Helpers & Utilities (2)

Type Summary
LossFunctionExtensions Tensor-shaped conveniences over ILossFunction, including the library’s single implementation of “gradient of a loss with respect to a prediction”.
ShapeContractExtensions Uniform access to the optional contract forms, whichever of them a contract actually implements.