Table of Contents

Enum GramSchmidtAlgorithmType

Namespace
AiDotNet.Enums.AlgorithmTypes
Assembly
AiDotNet.dll

Represents different algorithm types for the Gram-Schmidt orthogonalization process.

public enum GramSchmidtAlgorithmType

Fields

Classical = 0

Uses the Classical Gram-Schmidt algorithm for orthogonalization.

For Beginners: The Classical Gram-Schmidt algorithm is the original and most straightforward implementation of the orthogonalization process.

It works by taking each vector in sequence and subtracting from it any components that point in the same direction as the previously orthogonalized vectors. This leaves only the component that's perpendicular to all previous vectors.

Think of it like building a team where each new person brings only skills that nobody else on the team has yet. The first person brings all their skills, but the second person only contributes skills that the first person doesn't have, and so on.

The Classical method is simple to understand and implement. However, it can suffer from numerical instability when used with floating-point arithmetic on computers, especially for large sets of vectors. This means small rounding errors can accumulate and cause the vectors to not be perfectly orthogonal.

It's often suitable for educational purposes and for problems where high precision isn't critical or where the vectors are known to be well-behaved.

Modified = 1

Uses the Modified Gram-Schmidt algorithm for orthogonalization.

For Beginners: The Modified Gram-Schmidt algorithm is a more numerically stable version of the orthogonalization process.

While the Classical method computes all projections using the original vectors, the Modified method updates the vectors after each projection. This reduces the accumulation of rounding errors in computer calculations.

Imagine washing a very dirty cloth. The Classical method is like soaking the cloth once in clean water and expecting all the dirt to come out. The Modified method is like rinsing the cloth multiple times with fresh water after each rinse - it's more thorough and gets better results.

The Modified Gram-Schmidt algorithm is preferred in most practical applications, especially when:

  1. Working with large datasets

  2. High precision is required

  3. The vectors might be nearly linearly dependent (almost pointing in the same direction)

  4. The calculations are part of a larger algorithm where errors could propagate

Most professional software libraries use the Modified Gram-Schmidt algorithm by default because of its superior numerical properties.

Remarks

For Beginners: The Gram-Schmidt process is a method for converting a set of vectors into a set of orthogonal vectors (vectors that are perpendicular to each other).

Imagine you're in a room with walls that aren't at right angles to each other. The Gram-Schmidt process is like rearranging these walls so they're all perpendicular, making the room easier to measure and work with.

Why is this important in AI and machine learning?

  1. Feature Independence: In machine learning, we often want features that are independent of each other. Orthogonal vectors represent completely independent features, which can improve model performance.

  2. Dimensionality Reduction: Techniques like Principal Component Analysis (PCA) use orthogonalization to find the most important directions in your data.

  3. Numerical Stability: Many algorithms work better when using orthogonal vectors because calculations become more stable and accurate.

  4. Solving Systems of Equations: Orthogonal vectors make solving certain mathematical problems much easier.

The Gram-Schmidt process takes a set of vectors and, one by one, makes each new vector perpendicular to all previous vectors. This creates a new set of vectors that span the same space but are all perpendicular to each other.

This enum specifies which variation of the Gram-Schmidt algorithm to use, as there are different implementations with different numerical properties.

Edit this page