Table of Contents

Interface INormalizer<T, TInput, TOutput>

Namespace
AiDotNet.Interfaces
Assembly
AiDotNet.dll
public interface INormalizer<T, TInput, TOutput>

Type Parameters

T
TInput
TOutput

Methods

Denormalize(TInput, TOutput, TOutput, List<NormalizationParameters<T>>, NormalizationParameters<T>)

Calculates the denormalized Y-intercept (constant term) for a linear model.

T Denormalize(TInput xMatrix, TOutput y, TOutput coefficients, List<NormalizationParameters<T>> xParams, NormalizationParameters<T> yParams)

Parameters

xMatrix TInput

The original input feature matrix (before normalization).

y TOutput

The original target vector (before normalization).

coefficients TOutput

The model coefficients (can be either normalized or denormalized).

xParams List<NormalizationParameters<T>>

The normalization parameters used for the input features (X).

yParams NormalizationParameters<T>

The normalization parameters used for the target variable (Y).

Returns

T

The denormalized Y-intercept (constant term) for use with non-normalized data.

Remarks

This method computes the Y-intercept for a model trained on normalized data so it can be used with non-normalized input data.

For Beginners: This calculates the starting point (base value) for your model's predictions.

In a linear model like house price prediction:

  • Coefficients tell you how much each feature contributes (e.g., $100 per square foot)
  • The Y-intercept is the base value (e.g., $50,000 base price before adjustments)

When you train on normalized data:

  • The Y-intercept learned by the model works for normalized values
  • This method calculates the correct Y-intercept for original, non-normalized data

For example:

  • Your house price model might have a normalized Y-intercept of 0.5
  • This method converts it to a meaningful value like $150,000

This is typically used together with denormalized coefficients to create a complete model that works with original, non-normalized data.

Denormalize(TOutput, NormalizationParameters<T>)

Reverses the normalization of data using the original normalization parameters.

TOutput Denormalize(TOutput data, NormalizationParameters<T> parameters)

Parameters

data TOutput

The normalized data to denormalize.

parameters NormalizationParameters<T>

The normalization parameters that were used during normalization.

Returns

TOutput

The denormalized data in its original scale.

Remarks

This method transforms normalized data back to its original scale using the normalization parameters that were generated during the normalization process.

For Beginners: This converts normalized values back to their original scale.

For example:

  • If normalization converted ages (5, 18, 35, 62, 80) to (0.0, 0.17, 0.4, 0.76, 1.0)
  • Denormalization would convert (0.0, 0.17, 0.4, 0.76, 1.0) back to (5, 18, 35, 62, 80)

This is typically used:

  • After making predictions with a model (to convert predictions back to meaningful units)
  • When you need to interpret normalized data in its original context

Denormalize(TOutput, List<NormalizationParameters<T>>, NormalizationParameters<T>)

Denormalizes model coefficients to make them applicable to non-normalized input data.

TOutput Denormalize(TOutput coefficients, List<NormalizationParameters<T>> xParams, NormalizationParameters<T> yParams)

Parameters

coefficients TOutput

The model coefficients from a model trained on normalized data.

xParams List<NormalizationParameters<T>>

The normalization parameters used for the input features (X).

yParams NormalizationParameters<T>

The normalization parameters used for the target variable (Y).

Returns

TOutput

Denormalized coefficients that can be used with original, non-normalized data.

Remarks

This method adjusts the coefficients of a model trained on normalized data so they can be used directly with non-normalized input data.

For Beginners: This converts the model's internal numbers to work with your original data.

When you train a model on normalized data:

  • The model learns coefficients (weights) that work with normalized values
  • To use the model with original, non-normalized data, these coefficients need adjustment
  • This method performs that adjustment

For example:

  • You trained a house price model on normalized data (size, bedrooms, age)
  • The model learned coefficients like [0.7, 0.2, -0.1]
  • This method converts these to coefficients that work with the original units (e.g., dollars per square foot, dollars per bedroom, etc.)

This is typically used when:

  • You want to interpret what the model learned (e.g., "each additional bedroom adds $20,000")
  • You want to use the model with non-normalized input data

NormalizeInput(TInput)

Normalizes input data to a standard range.

(TInput, List<NormalizationParameters<T>>) NormalizeInput(TInput data)

Parameters

data TInput

The data to normalize.

Returns

(TInput, List<NormalizationParameters<T>>)

A tuple containing:

  • The normalized data
  • A list of normalization parameters for each feature (needed for later denormalization)

Remarks

This method transforms each feature of the input data to a normalized scale and returns both the normalized data and the parameters used for normalization of each feature.

For Beginners: This converts a table of numbers to a standard scale.

For example:

  • If your data contains information about houses (columns for size, bedrooms, age, price), each feature is normalized separately
  • Size values might range from 1,000-5,000 sq ft, but are converted to 0-1
  • Bedroom values might range from 1-6, but are also converted to 0-1
  • The normalization parameters for each feature are saved separately

This is typically used to prepare multiple features (input variables) for machine learning.

NormalizeOutput(TOutput)

Normalizes output data to a standard range.

(TOutput, NormalizationParameters<T>) NormalizeOutput(TOutput data)

Parameters

data TOutput

The data to normalize.

Returns

(TOutput, NormalizationParameters<T>)

A tuple containing:

  • The normalized data
  • The normalization parameters that were used (needed for later denormalization)

Remarks

This method transforms the input data to a normalized scale and returns both the normalized data and the parameters used for normalization.

For Beginners: This converts a list of numbers to a standard scale.

For example:

  • If your data contains ages (5, 18, 35, 62, 80), normalization might convert these to values between 0 and 1: (0.0, 0.17, 0.4, 0.76, 1.0)
  • The normalization parameters (like minimum and maximum values) are also returned so you can reverse this process later

This is typically used to prepare a single feature or target variable for machine learning.

Edit this page