Projections

Projections explain how one vector can be expressed as a shadow or component along another vector or subspace.

This idea connects geometry, algebra, optimization, and machine learning. Whenever we approximate, predict, or minimize error, we are using projections.

---

Why Projections Are Important

In real life, we often need the “best approximation” rather than an exact value.

Projections help us:

• Approximate vectors
• Remove unwanted components
• Minimize error
• Understand orthogonality

Many AI algorithms are projection-based.

---

What Is a Projection? (Intuition)

Imagine shining a light on a vector.

The shadow cast on another direction is the projection.

It tells us how much of one vector lies along another.

---

Projection Onto a Vector

Given two vectors:

• v → vector being projected
• u → direction vector

The projection of v onto u is the component of v that lies along u.

---

Formula for Projection Onto a Vector

The projection of v onto u is:

proj_u(v) = ( v·u / u·u ) u

This formula is extremely important for exams and applications.

---

Understanding the Projection Formula

Each part of the formula has meaning:

• v·u → measures alignment
• u·u → magnitude squared of u
• Multiplying by u → gives direction

The result is a vector along u.

---

Example: Projection Onto a Vector

Let:

v = (3, 4), u = (1, 0)

Then:

v·u = 3
u·u = 1

proj_u(v) = 3(1, 0) = (3, 0)

This is the horizontal component of v.

---

Orthogonal Component

The part of v that is not along u is called the orthogonal component.

It is given by:

v − proj_u(v)

This component is perpendicular to u.

---

Orthogonal Projection (Key Idea)

An orthogonal projection is the closest vector (in distance) to v that lies along u.

This makes projections essential for minimizing error.

---

Projection Onto a Line

Projecting onto a vector is the same as projecting onto the line spanned by that vector.

The projection gives the closest point on that line to the original vector.

---

Projection Onto a Subspace

Projections can also be done onto entire subspaces.

If a subspace has an orthonormal basis, projection becomes very simple.

Each basis vector contributes independently.

---

Orthogonal Basis and Projections

If basis vectors are orthogonal:

• Projections are easy to compute
• No interference between directions

This is why orthogonality is so powerful.

---

Projection in Geometry

Geometrically, projections represent:

• Shadows
• Perpendicular drops
• Closest points

They appear naturally in shapes and measurements.

---

Projection in Physics

Physics uses projections constantly:

• Resolving forces
• Motion along inclined planes
• Component analysis

Each force is projected onto directions.

---

Projection in Data Science

In data science:

• Data is projected onto features
• Noise is removed using projections
• Dimensionality is reduced

PCA is fundamentally projection-based.

---

Projection in Machine Learning

Machine learning uses projections in:

• Linear regression
• Least squares problems
• Feature extraction

Predictions are projections onto model spaces.

---

Least Squares and Projections

Least squares finds the vector in a subspace that is closest to a given data vector.

That closest vector is an orthogonal projection.

This is a core ML idea.

---

Projections in Competitive Exams

Exams often test:

• Projection formula
• Geometric interpretation
• Orthogonality conditions

Understanding beats memorization here.

---

Common Mistakes to Avoid

Students often make these mistakes:

• Forgetting dot product in formula
• Projecting onto wrong vector
• Confusing projection with component subtraction

Always check direction and formula.

---

Practice Questions

---

Quick Quiz

---

Quick Recap

• Projections measure components along directions
• Computed using dot products
• Orthogonal projection minimizes distance
• Core idea in least squares and ML
• Connects geometry to data science

With projections mastered, you are now ready to learn orthogonality, which explains perpendicular structure in spaces.