Table of Contents

Enum GradientType

Namespace
AiDotNet.Enums
Assembly
AiDotNet.dll

Specifies different types of gradient descent optimization algorithms used in machine learning.

public enum GradientType

Fields

Alpha = 1

Represents a gradient descent algorithm that prioritizes speed of convergence.

For Beginners: Alpha gradient is designed for speed, taking larger steps to reach the solution faster.

Think of it like a sports car that:

  • Moves quickly toward the destination
  • Can reach the solution in fewer iterations
  • Works well when the "landscape" is smooth
  • Might overshoot or become unstable on tricky problems

Alpha gradient is good when:

  • You need quick results
  • Your data is well-behaved (not too noisy)
  • You're willing to risk some instability for speed
Beta = 2

Represents a gradient descent algorithm that prioritizes stability and reliability over speed.

For Beginners: Beta gradient is designed for reliability, taking careful steps to ensure stable learning.

Think of it like an off-road vehicle that:

  • Moves more cautiously but consistently
  • Is less likely to get stuck or go off-track
  • Handles difficult or noisy data better
  • Takes more time but is more dependable

Beta gradient is good when:

  • Your data is noisy or complex
  • You value consistent, reliable results
  • You're willing to wait longer for a solution
  • You're dealing with a challenging problem where other methods might fail
Omega = 0

Represents the standard gradient descent algorithm with momentum and adaptive learning rates.

For Beginners: Omega gradient combines the best features of several optimization methods.

Think of it like a smart car that:

  • Remembers which direction was working well (momentum)
  • Automatically adjusts its speed based on the terrain (adaptive learning)
  • Can handle both steep and gentle slopes efficiently

This is often a good default choice when you're not sure which gradient type to use, as it balances speed and stability for most common machine learning problems.

Remarks

For Beginners: Gradient descent is how AI models learn from data. Think of it like finding the lowest point in a valley by taking steps downhill. Different gradient types represent different strategies for how to take these steps - some are faster but riskier, others are slower but more reliable.

Edit this page