Class ModelMetadata<T>

Namespace

AiDotNet.Models

Assembly

AiDotNet.dll

Represents metadata about a machine learning model, including its type, complexity, and additional descriptive information.

public class ModelMetadata<T>

Type Parameters

T

The numeric type used for calculations, typically float or double.

Inheritance

object

ModelMetadata<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This class encapsulates metadata about a machine learning model, providing information that describes the model's characteristics without containing the actual model implementation. It includes details such as the model type, feature count, complexity, and a textual description. Additionally, it provides an extensible dictionary for storing arbitrary additional information and can store serialized model data. This metadata is useful for model cataloging, selection, and management purposes.

For Beginners: This class stores descriptive information about a machine learning model.

When working with machine learning models:

• You often need to know what type of model it is (linear regression, neural network, etc.)
• You want to understand its complexity and what features it uses
• You may need to store additional information about how it was created or should be used

This class stores all that information, including:

• The type of model (classification, regression, etc.)
• How many features (input variables) it uses
• How complex the model is
• A human-readable description
• Any additional custom information you want to include
• The actual model data in serialized form

This metadata helps you understand what a model does and how it works without having to examine the model itself.

Properties

AdditionalInfo

Gets or sets additional information about the model as key-value pairs.

public Dictionary<string, object> AdditionalInfo { get; set; }

Property Value

Dictionary<string, object>

A dictionary containing additional model information.

Remarks

This property provides an extensible way to store additional information about the model as key-value pairs. This can include any information that doesn't fit into the other properties, such as the date the model was created, the author, the dataset used for training, performance metrics, hyperparameter values, or any other relevant metadata. The flexible dictionary structure allows for storing arbitrary information without modifying the class structure.

For Beginners: This allows you to store any additional custom information about the model.

The additional info dictionary:

• Stores any extra information as key-value pairs
• Is flexible and can contain any type of data
• Allows you to extend the metadata without changing the class structure

Common types of information you might store:

• Creation date: When the model was built
• Author: Who created the model
• Training dataset: What data was used to train it
• Performance metrics: How well it performed in testing
• Hyperparameters: Specific configuration settings used

This flexibility is useful because:

• Different models may need different types of metadata
• You can add custom information without modifying the class
• It allows for future extensibility as needs change

Complexity

Gets or sets a measure of the model's complexity.

public int Complexity { get; set; }

Property Value

int

An integer representing the model's complexity.

Remarks

This property provides a measure of the model's complexity, which can vary depending on the type of model. For example, in a linear model, complexity might refer to the number of coefficients; in a decision tree, it might refer to the depth or number of nodes; in a neural network, it might refer to the number of layers or parameters. Higher complexity can allow a model to capture more intricate patterns but may also increase the risk of overfitting and the computational resources required.

For Beginners: This indicates how complex or sophisticated the model is.

The complexity measure:

• Gives you an idea of how intricate the model's structure is
• Can mean different things for different model types
• Higher values generally indicate more complex models

For different model types, complexity might represent:

• Linear models: Number of coefficients or terms
• Decision trees: Depth of the tree or number of nodes
• Neural networks: Number of layers or parameters

This information is useful because:

• More complex models can capture more intricate patterns
• But they may also be more prone to overfitting
• And they typically require more computational resources

Description

Gets or sets a human-readable description of the model.

public string Description { get; set; }

Property Value

string

A string containing a description of the model.

Remarks

This property provides a human-readable description of the model, which can include information about its purpose, how it was created, its strengths and limitations, or any other relevant details. The description is useful for documentation purposes and for helping users understand what the model does and how it should be used.

For Beginners: This provides a human-readable explanation of what the model does and how it works.

The description:

• Explains the model's purpose and functionality in plain language
• May include information about how it was created or trained
• Can describe the model's strengths, limitations, or intended use cases

For example, a description might be: "This is a gradient boosting model trained to predict house prices based on 10 features including square footage, location, and number of bedrooms. It was trained on data from 2010-2020 and performs best on properties in urban areas."

This information is valuable for:

• Documentation purposes
• Helping others understand the model
• Remembering the model's purpose and limitations later

FeatureCount

Gets or sets the number of features used by the model.

public int FeatureCount { get; set; }

Property Value

int

An integer representing the number of input features.

Remarks

This property indicates how many input features or variables the model uses to make predictions. The feature count is an important aspect of the model's structure and can affect its performance, interpretability, and computational requirements. Models with more features may capture more complex patterns but might also be more prone to overfitting and require more data to train effectively.

For Beginners: This tells you how many input variables the model uses to make predictions.

The feature count:

• Indicates how many different input variables the model considers
• Affects the model's complexity and interpretability
• Influences how much data is needed to train the model effectively

For example, a house price prediction model might use 10 features such as square footage, number of bedrooms, location, etc.

This information is important because:

• It helps you prepare the right number of inputs when using the model
• It gives you an idea of the model's complexity
• It can indicate potential issues (too many features might cause overfitting)

FeatureImportance

Gets or sets the importance of each feature in the model.

public Dictionary<string, T> FeatureImportance { get; set; }

Property Value

Dictionary<string, T>

A dictionary mapping feature names to their importance values.

Remarks

This property stores the importance of each feature used by the model. Feature importance indicates how much each input variable contributes to the model's predictions. The exact meaning and scale of importance values can vary depending on the model type and the method used to calculate importance. This information is crucial for understanding which features have the most significant impact on the model's output and can be used for feature selection or model interpretation.

For Beginners: This shows how important each input variable is to the model's predictions.

The feature importance dictionary:

• Maps each feature (input variable) name to a number representing its importance
• Higher values typically indicate more important features
• The exact meaning of the values depends on the type of model and how importance was calculated

This is useful for:

• Understanding which inputs have the biggest impact on predictions
• Identifying irrelevant or less important features
• Simplifying the model by focusing on the most important features

For example, in a house price prediction model, you might see that 'location' has a high importance value, while 'roof color' has a low value, indicating that location strongly influences the prediction but roof color doesn't.

ModelData

Gets or sets the serialized model data.

public byte[] ModelData { get; set; }

Property Value

byte[]

A byte array containing the serialized model data.

Remarks

This property stores the serialized model data as a byte array. This can be used to store a compact representation of the model itself, allowing the model to be reconstructed from the metadata. The serialization format depends on the specific model implementation and serialization mechanism used. This property is particularly useful for model persistence, sharing, and deployment scenarios.

For Beginners: This stores the actual model in a serialized (binary) format.

The model data:

• Contains the serialized (converted to bytes) model itself
• Allows the model to be stored, transmitted, or reconstructed later
• Is in a format that depends on how the model was serialized

This is useful for:

• Saving the model to a file or database
• Transmitting the model over a network
• Reconstructing the model without having to retrain it

For example, you might serialize a trained model to this property, then save the entire ModelMetadata object to a file. Later, you can load the file and deserialize the model data to recreate the working model.

ModelType

Gets or sets the type of the model.

public ModelType ModelType { get; set; }

Property Value

ModelType

A ModelType enumeration value indicating the model's type.

Remarks

This property indicates the type of the model, such as regression, classification, clustering, or time series. The model type provides a high-level categorization of what the model does and what kind of problems it is designed to solve. This information is useful for understanding the model's purpose and for selecting appropriate models for specific tasks.

For Beginners: This tells you what kind of problem the model is designed to solve.

The model type:

• Indicates whether the model is for regression, classification, clustering, etc.
• Helps you understand what the model is designed to do
• Guides how the model's outputs should be interpreted

Common model types include:

• Regression: Predicts continuous values (like prices, temperatures)
• Classification: Predicts categories or classes (like spam/not spam)
• Clustering: Groups similar items together
• Time Series: Makes predictions based on time-ordered data

Knowing the model type is essential for using the model correctly and understanding what kind of predictions it can make.

Name

Gets or sets the name of the model.

public string Name { get; set; }

Property Value

string

A string representing the model's name.

Remarks

This property provides a human-readable name for the model, useful for identification and cataloging purposes.

Properties

Gets custom properties associated with the model.

public Dictionary<string, object> Properties { get; }

Property Value

Dictionary<string, object>

A dictionary containing custom properties as key-value pairs.

Remarks

This property provides an extensible way to store custom properties and configuration settings specific to the model. It complements the AdditionalInfo property by providing a dedicated space for model-specific properties. Use SetProperty(string, object) and RemoveProperty(string) methods to modify the properties.

TrainingDate

Gets or sets the date and time (with timezone) when the model was trained.

public DateTimeOffset? TrainingDate { get; set; }

Property Value

DateTimeOffset?

A nullable DateTimeOffset representing when the model was trained, or null if unknown.

Remarks

This property stores the date and time (including timezone information) when the model was trained. It is nullable, allowing you to indicate when the training date is unknown or not set. Using DateTimeOffset ensures accurate tracking across different time zones.

Version

Gets or sets the version of the model.

public string Version { get; set; }

Property Value

string

A string representing the model's version.

Remarks

This property indicates the version of the model, which can be useful for tracking changes and updates over time.

Methods

RemoveProperty(string)

Removes a custom property from the Properties dictionary.

public bool RemoveProperty(string key)

Parameters

key string

The property key to remove.

Returns

bool

True if the property was removed; otherwise, false.

SetProperty(string, object)

Adds or updates a custom property in the Properties dictionary.

public void SetProperty(string key, object value)

Parameters

key string

The property key.

value object

The property value.