Class MAEConfig

Namespace

AiDotNet.SelfSupervisedLearning

Assembly

AiDotNet.dll

MAE-specific configuration settings.

public class MAEConfig

Inheritance

object

MAEConfig

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

For Beginners: MAE (Masked Autoencoder) learns by masking random patches of an image and training the model to reconstruct them.

Properties

DecoderDepth

Gets or sets the number of decoder transformer blocks.

public int? DecoderDepth { get; set; }

Property Value

int?

Remarks

Default: 8

DecoderEmbedDimension

Gets or sets the decoder embedding dimension.

public int? DecoderEmbedDimension { get; set; }

Property Value

int?

Remarks

Default: 512

Decoder is typically smaller than encoder.

DecoderNumHeads

Gets or sets the number of decoder attention heads.

public int? DecoderNumHeads { get; set; }

Property Value

int?

Remarks

Default: 16

MaskRatio

Gets or sets the fraction of patches to mask.

public double? MaskRatio { get; set; }

Property Value

double?

Remarks

Default: 0.75

MAE uses aggressive masking (75%) which is efficient and effective.

NormalizeTarget

Gets or sets whether to normalize reconstruction target.

public bool? NormalizeTarget { get; set; }

Property Value

bool?

Remarks

Default: true

Normalizing patches before reconstruction typically helps.

PatchSize

Gets or sets the patch size.

public int? PatchSize { get; set; }

Property Value

int?

Remarks

Default: 16

Common sizes are 16 (for 224x224 images, giving 196 patches) or 14 (for ViT).

Methods

GetConfiguration()

Gets the configuration as a dictionary.

public IDictionary<string, object> GetConfiguration()

Returns

IDictionary<string, object>