Table of Contents

Class NASResultSummary

Namespace
AiDotNet.Models.Results
Assembly
AiDotNet.dll

Represents a redacted summary of Neural Architecture Search (NAS) results.

public sealed class NASResultSummary
Inheritance
NASResultSummary
Inherited Members

Remarks

This summary captures the discovered architecture and any hardware-aware optimizations without exposing proprietary implementation details.

For Beginners: After NAS completes, this tells you what architecture was discovered: - How many layers/nodes were selected - What operations were chosen at each position - Any hardware constraints that were considered

Properties

ArchitectureDescription

Gets or sets a human-readable description of the discovered architecture.

public string? ArchitectureDescription { get; set; }

Property Value

string

Remarks

Example: "MobileNetV3-like with 12 layers, 3 skip connections, avg pooling"

DiscoveredNodeCount

Gets or sets the number of nodes/layers in the discovered architecture.

public int? DiscoveredNodeCount { get; set; }

Property Value

int?

EstimatedFLOPs

Gets or sets the estimated FLOPs (floating-point operations) of the discovered architecture.

public long? EstimatedFLOPs { get; set; }

Property Value

long?

EstimatedParameters

Gets or sets the estimated parameter count of the discovered architecture.

public long? EstimatedParameters { get; set; }

Property Value

long?

FinalArchitectureScore

Gets or sets the final architecture score (validation accuracy or combined metric).

public double? FinalArchitectureScore { get; set; }

Property Value

double?

LatencyConstraintMs

Gets or sets the latency constraint in milliseconds, if specified.

public double? LatencyConstraintMs { get; set; }

Property Value

double?

MemoryConstraintMB

Gets or sets the memory constraint in megabytes, if specified.

public double? MemoryConstraintMB { get; set; }

Property Value

double?

OFASubnet

Gets or sets OnceForAll-specific subnet information when OFA strategy was used.

public OFASubnetSummary? OFASubnet { get; set; }

Property Value

OFASubnetSummary

Remarks

OnceForAll trains a single "supernet" and then extracts specialized subnets. This field describes the extracted subnet configuration.

QuantizationAware

Gets or sets whether quantization-aware search was enabled.

public bool QuantizationAware { get; set; }

Property Value

bool

SearchIterations

Gets or sets the number of architecture search epochs/iterations completed.

public int? SearchIterations { get; set; }

Property Value

int?

SelectedOperations

Gets or sets a list of operation types selected at each position in the architecture.

public List<string>? SelectedOperations { get; set; }

Property Value

List<string>

Remarks

This provides a high-level view of the discovered architecture without exposing exact weights or proprietary details.

Example: ["conv3x3", "skip", "sep_conv5x5", "max_pool", "conv3x3"]

TargetPlatform

Gets or sets the target hardware platform, if hardware-aware search was used.

public string? TargetPlatform { get; set; }

Property Value

string