Class CNAPOptions<T, TInput, TOutput>

Namespace

AiDotNet.MetaLearning.Options

Assembly

AiDotNet.dll

Configuration options for the Conditional Neural Adaptive Processes (CNAP) algorithm.

public class CNAPOptions<T, TInput, TOutput> : IMetaLearnerOptions<T>

Type Parameters

T

The numeric data type used for calculations (e.g., float, double).

TInput

The input data type (e.g., Matrix<T>, Tensor<T>).

TOutput

The output data type (e.g., Vector<T>, Tensor<T>).

Inheritance

object

CNAPOptions<T, TInput, TOutput>

Implements

IMetaLearnerOptions<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

CNAP extends Neural Processes by conditioning on task-specific context points and learning to produce fast adaptation weights for each task. This enables effective few-shot learning through learned task representations.

For Beginners: CNAP learns to understand tasks from examples: 1. It encodes examples (context points) into a task representation 2. It uses this representation to generate task-specific fast weights 3. The fast weights modify the base model to work well on that specific task

This is like having a teacher who can quickly understand what kind of problem you're working on and adjust their teaching style accordingly.

Key Components: - Encoder: Processes context points into task representations - Adaptation Network: Generates task-specific fast weights from representations - Base Model: The neural network that gets modified by fast weights

Constructors

CNAPOptions(IFullModel<T, TInput, TOutput>)

Initializes a new instance of the CNAPOptions class with the required meta-model.

public CNAPOptions(IFullModel<T, TInput, TOutput> metaModel)

Parameters

metaModel IFullModel<T, TInput, TOutput>

The meta-model to be trained (required).

Exceptions

ArgumentNullException

Thrown when metaModel is null.

Properties

AdaptationNetworkLayers

Gets or sets the number of layers in the adaptation network.

public int AdaptationNetworkLayers { get; set; }

Property Value

int

Default: 2.

AdaptationSteps

Gets or sets the number of gradient steps to take during inner loop adaptation.

public int AdaptationSteps { get; set; }

Property Value

int

Default: 1 (CNAP typically uses single-step adaptation with fast weights).

CheckpointFrequency

Gets or sets how often to save checkpoints.

public int CheckpointFrequency { get; set; }

Property Value

int

Default: 1000.

ContextSize

Gets or sets the number of context points to use for encoding (support set size).

public int ContextSize { get; set; }

Property Value

int

Default: 5 (for 5-shot learning).

DataLoader

Gets or sets the episodic data loader for sampling tasks.

public IEpisodicDataLoader<T, TInput, TOutput>? DataLoader { get; set; }

Property Value

IEpisodicDataLoader<T, TInput, TOutput>

Default: null (tasks must be provided manually to MetaTrain).

DropoutRate

Gets or sets the dropout rate for regularization.

public double DropoutRate { get; set; }

Property Value

double

Default: 0.1.

EnableCheckpointing

Gets or sets whether to save checkpoints during training.

public bool EnableCheckpointing { get; set; }

Property Value

bool

Default: false.

EncoderLayers

Gets or sets the number of layers in the encoder network.

public int EncoderLayers { get; set; }

Property Value

int

Default: 3.

EvaluationFrequency

Gets or sets how often to evaluate during meta-training.

public int EvaluationFrequency { get; set; }

Property Value

int

Default: 500.

EvaluationTasks

Gets or sets the number of tasks to use for evaluation.

public int EvaluationTasks { get; set; }

Property Value

int

Default: 100.

FastWeightMode

Gets or sets how fast weights are applied to the model.

public FastWeightApplicationMode FastWeightMode { get; set; }

Property Value

FastWeightApplicationMode

Default: Additive.

Remarks

- Additive: Fast weights are added to base weights (θ + α) - Multiplicative: Fast weights scale base weights (θ × α) - FiLM: Feature-wise Linear Modulation (γθ + β)

FastWeightRegularization

Gets or sets the regularization weight for fast weights.

public double FastWeightRegularization { get; set; }

Property Value

double

Default: 0.001.

FastWeightScale

Gets or sets the scale factor for fast weights.

public double FastWeightScale { get; set; }

Property Value

double

Default: 1.0.

GradientClipThreshold

Gets or sets the maximum gradient norm for gradient clipping.

public double? GradientClipThreshold { get; set; }

Property Value

double?

Default: 10.0.

HiddenDimension

Gets or sets the hidden dimension for the encoder and adaptation networks.

public int HiddenDimension { get; set; }

Property Value

int

Default: 512.

InnerLearningRate

Gets or sets the learning rate for the inner loop (task adaptation).

public double InnerLearningRate { get; set; }

Property Value

double

Default: 0.01.

InnerOptimizer

Gets or sets the optimizer for inner-loop adaptation.

public IGradientBasedOptimizer<T, TInput, TOutput>? InnerOptimizer { get; set; }

Property Value

IGradientBasedOptimizer<T, TInput, TOutput>

Default: null (uses built-in SGD optimizer).

LossFunction

Gets or sets the loss function for training.

public ILossFunction<T>? LossFunction { get; set; }

Property Value

ILossFunction<T>

Default: null (uses model's default loss function if available).

MetaBatchSize

Gets or sets the number of tasks to sample per meta-training iteration.

public int MetaBatchSize { get; set; }

Property Value

int

Default: 4.

MetaModel

Gets or sets the meta-model to be trained. This is the only required property.

public IFullModel<T, TInput, TOutput> MetaModel { get; set; }

Property Value

IFullModel<T, TInput, TOutput>

Remarks

The meta-model is the base network that will be modified by fast weights during task adaptation. It should be a neural network with parameters that can be efficiently modulated.

MetaOptimizer

Gets or sets the optimizer for meta-parameter updates (outer loop).

public IGradientBasedOptimizer<T, TInput, TOutput>? MetaOptimizer { get; set; }

Property Value

IGradientBasedOptimizer<T, TInput, TOutput>

Default: null (uses built-in Adam optimizer).

NormalizeFastWeights

Gets or sets whether to normalize fast weights.

public bool NormalizeFastWeights { get; set; }

Property Value

bool

Default: true.

Remarks

Normalizing fast weights helps stabilize training and prevents the adaptation from making overly large changes to the base model.

NumAttentionHeads

Gets or sets the number of attention heads when using attention.

public int NumAttentionHeads { get; set; }

Property Value

int

Default: 4.

NumMetaIterations

Gets or sets the total number of meta-training iterations to perform.

public int NumMetaIterations { get; set; }

Property Value

int

Default: 10000.

OuterLearningRate

Gets or sets the learning rate for the outer loop (meta-update).

public double OuterLearningRate { get; set; }

Property Value

double

Default: 0.001.

PredictUncertainty

Gets or sets whether to predict uncertainty along with predictions.

public bool PredictUncertainty { get; set; }

Property Value

bool

Default: false.

RandomSeed

Gets or sets the random seed for reproducibility.

public int? RandomSeed { get; set; }

Property Value

int?

Default: null (non-deterministic).

RepresentationDimension

Gets or sets the dimension of the task representation vector.

public int RepresentationDimension { get; set; }

Property Value

int

Default: 256.

Remarks

The representation dimension controls the capacity of the task encoding. Higher values can capture more complex task relationships but require more computation and may be prone to overfitting.

UncertaintyWeight

Gets or sets the weight for uncertainty loss when predicting uncertainty.

public double UncertaintyWeight { get; set; }

Property Value

double

Default: 0.1.

UseAttention

Gets or sets whether to use attention for aggregating context points.

public bool UseAttention { get; set; }

Property Value

bool

Default: true.

Remarks

When enabled, the encoder uses self-attention to aggregate information from multiple context points. This allows the model to learn which context points are most relevant for each task.

For Beginners: Attention helps the model focus on the most important examples when understanding a new task, rather than treating all examples equally.

UseFirstOrder

Gets or sets whether to use first-order approximation.

public bool UseFirstOrder { get; set; }

Property Value

bool

Default: true (CNAP typically uses first-order for efficiency).

UseLayerNorm

Gets or sets whether to use layer normalization in networks.

public bool UseLayerNorm { get; set; }

Property Value

bool

Default: true.

Methods

Clone()

Creates a deep copy of the CNAP options.

public IMetaLearnerOptions<T> Clone()

Returns

IMetaLearnerOptions<T>

A new CNAPOptions instance with the same configuration values.

IsValid()

Validates that all CNAP configuration options are properly set.

public bool IsValid()

Returns

bool

True if the configuration is valid for CNAP training; otherwise, false.