15) Vector Norms
Vector Norms: Measuring Vector Magnitude
1. What is a Vector Norm?
A vector norm is a function that assigns a positive length to a vector, measuring its size.
Real-World Meaning
- Physics: The norm of velocity gives the speed of an object.
- Machine Learning: Norms measure distances between data points.
- Optimization: Used in gradient descent to control updates.
2. Types of Vector Norms
(a) Euclidean Norm (L2 Norm)
Measures straight-line distance.
||v||₂ = sqrt(v₁² + v₂² + ... + vₙ²)
Real-World Use:
- GPS Navigation: Finding shortest distance.
- Physics: Computing magnitudes of forces.
(b) Manhattan Norm (L1 Norm)
Sum of absolute values of vector components.
||v||₁ = |v₁| + |v₂| + ... + |vₙ|
Real-World Use:
- City Blocks: Distance in grid-like street layouts.
- AI: Used in Lasso Regression for feature selection.
(c) Max Norm (L∞ Norm)
Largest absolute value of any vector component.
||v||∞ = max(|v₁|, |v₂|, ..., |vₙ|)
Real-World Use:
- Quality Control: Finding maximum deviation.
- Robotics: Avoiding worst-case movement errors.
(d) p-Norm (Lp Norm)
Generalized norm formula.
||v||ₚ = (∑ |vᵢ|ᵖ)^(1/p)
Real-World Use:
- Deep Learning: Used in error functions.
- Image Processing: Measures pixel differences.
3. Properties of Norms
- Non-negativity: ||v|| ≥ 0
- Zero Vector Property: ||v|| = 0 ⇔ v = 0
- Triangle Inequality: ||u + v|| ≤ ||u|| + ||v||
- Scaling: ||c v|| = |c| ||v||
Real-World Meaning:
- Triangle inequality ensures the shortest path is direct.
- Scaling means doubling force doubles its effect.
4. Choosing the Right Norm
- L2 norm: Used when true distance matters.
- L1 norm: Used in grid-like systems.
- L∞ norm: Used when worst-case scenarios matter.
- Lp norm: Used in generalized settings.
Comments
Post a Comment