Class DenseNetConfiguration

Namespace

AiDotNet.Configuration

Assembly

AiDotNet.dll

Configuration options for DenseNet neural network architectures.

public class DenseNetConfiguration

Inheritance

object

DenseNetConfiguration

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

DenseNet (Densely Connected Convolutional Networks) connects each layer to every other layer in a feed-forward fashion, enabling strong gradient flow and feature reuse.

For Beginners: DenseNet is designed to maximize information flow by connecting each layer directly to all subsequent layers. This configuration lets you choose which variant to use and customize parameters like growth rate and compression factor.

Constructors

DenseNetConfiguration(DenseNetVariant, int, int, int, int, int, double, int[]?)

Initializes a new instance of the DenseNetConfiguration class.

public DenseNetConfiguration(DenseNetVariant variant, int numClasses, int inputHeight = 224, int inputWidth = 224, int inputChannels = 3, int growthRate = 32, double compressionFactor = 0.5, int[]? customBlockLayers = null)

Parameters

variant DenseNetVariant

The DenseNet variant to use.

numClasses int

The number of output classes for classification.

inputHeight int

The height of input images (default: 224).

inputWidth int

The width of input images (default: 224).

inputChannels int

The number of input channels (default: 3 for RGB).

growthRate int

The growth rate (default: 32).

compressionFactor double

The compression factor for transition layers (default: 0.5).

customBlockLayers int[]

Custom block layers (required when variant is Custom).

Properties

CompressionFactor

Gets the compression factor for transition layers.

public double CompressionFactor { get; }

Property Value

double

Remarks

For Beginners: Compression factor (theta) controls channel reduction at transition layers. A value of 0.5 means halving the channels at each transition, which helps control model size.

CustomBlockLayers

Gets the custom block layers configuration (only used when Variant is Custom).

public int[]? CustomBlockLayers { get; }

Property Value

int[]

GrowthRate

Gets the growth rate (k in the paper).

public int GrowthRate { get; }

Property Value

int

Remarks

For Beginners: The growth rate determines how many new feature maps each layer adds. Typical values are 12, 24, or 32. Higher values increase capacity but also computational cost.

InputChannels

Gets the number of input channels.

public int InputChannels { get; }

Property Value

int

InputHeight

Gets the height of input images in pixels.

public int InputHeight { get; }

Property Value

int

InputShape

Gets the computed input shape as [channels, height, width].

public int[] InputShape { get; }

Property Value

int[]

InputWidth

Gets the width of input images in pixels.

public int InputWidth { get; }

Property Value

int

NumClasses

Gets the number of output classes for classification.

public int NumClasses { get; }

Property Value

int

Variant

Gets the DenseNet variant to use.

public DenseNetVariant Variant { get; }

Property Value

DenseNetVariant

Methods

CreateDenseNet121(int)

Creates a DenseNet-121 configuration (recommended default).

public static DenseNetConfiguration CreateDenseNet121(int numClasses)

Parameters

numClasses int

Returns

DenseNetConfiguration

CreateForTesting(int)

Creates a minimal DenseNet configuration optimized for fast test execution.

public static DenseNetConfiguration CreateForTesting(int numClasses)

Parameters

numClasses int

The number of output classes.

Returns

DenseNetConfiguration

A minimal DenseNet configuration for testing.

Remarks

Uses [2, 2, 2, 2] block configuration with small growth rate (8) and 32x32 input, resulting in approximately 8 dense layers instead of 58+ in DenseNet-121. Construction time is typically under 50ms.

GetBlockLayers()

Gets the number of layers per dense block for this variant.

public int[] GetBlockLayers()

Returns

int[]

GetExpectedLayerCount()

Gets the expected total layer count for this configuration without constructing the network.

public int GetExpectedLayerCount()

Returns

int

The expected number of layers in the network.

Remarks

This is useful for tests that need to compare layer counts without the overhead of actually constructing the networks. Formula: 1 (stem conv) + 1 (stem BN) + 1 (stem pool) + sum(block_layers * 2) for BN+Conv in each dense layer + (num_blocks - 1) * 2 for transition layers (Conv + Pool) + 1 (final BN) + 1 (classifier)