Vector Operations: Navigating Space

1. Introduction

A vector is not just a static list of numbers; it describes a movement in space. If a vector represents “Walk 5 meters North,” then vector operations allow us to answer questions like:

  • “If I walk North, then East, where do I end up?” (Addition)
  • “If I walk twice as far?” (Scalar Multiplication)
  • “What is the path from A to B?” (Subtraction)

In Machine Learning, these operations are the engine of learning. When a model “learns,” it is literally moving its weight vector down a hill (Gradient Descent) towards a better solution.


2. Vector Addition (The Head-to-Tail Rule)

When adding two vectors u and v, we place the tail of v at the head of u. The result (sum) is the vector from the start of u to the end of v.

  • u = [ 2, 1 ]T
  • v = [ 1, 3 ]T

u + v = [ 2+1, 1+3 ]T = [ 3, 4 ]T

[!TIP] ML Application: In Gradient Descent, we update our current position (weights) by adding a step vector (the negative gradient). new_weights = old_weights + (-learning_rate * gradient)


3. Vector Subtraction

Subtracting a vector is adding its negative: u - v = u + (-v). Geometrically, u - v is the vector that points from v to u.

u - v = [ 2-1, 1-3 ]T = [ 1, -2 ]T

[!TIP] ML Application: This is used to calculate Error or Loss. If y is the prediction and t is the target (truth), the error vector is y - t. We want to make this vector as small as possible.


4. Scalar Multiplication

Multiplying a vector by a Scalar (single number) scales its magnitude without changing its direction (unless the scalar is negative, which flips it).

2 × [ 2, 1 ]T = [ 4, 2 ]T

This “stretches” or “shrinks” the vector. In ML, the Learning Rate acts as a scalar multiplier, controlling how big of a step we take.


5. Interactive Visualizer: Vector Playground

Drag the Blue (u) and Green (v) vector heads to see how operations change the Resultant Red (r) vector.

u: [0,0]
v: [0,0]
Result: [0,0]
Tip: Drag the Blue/Green dots.

6. Summary

  • Addition: Combining forces (e.g., Wind + Velocity).
  • Subtraction: Finding the difference or path between two points.
  • Scalar Multiplication: Scaling the magnitude without changing direction.

Next: Matrix Multiplication →