Linear Algebra
10.98 min read

Continuity

A function ff is continuous at x0x_0 if three things hold: (1) ff is defined at x0x_0; (2) the limit limxx0f(x)\lim_{x \to x_0} f(x) exists; and (3) the limit equals the function value: limxx0f(x)=f(x0)\lim_{x \to x_0} f(x) = f(x_0). Intuitively, continuity means you can draw the graph without lifting your pen.

Continuity fails in three ways: a removable discontinuity (limit exists but f(x0)f(x_0) is wrong), a jump discontinuity (left and right limits exist but differ), or an infinite/oscillating discontinuity (limit does not exist). A function is continuous on an interval if it is continuous at every point in that interval.

The class of continuous functions is called C0C^0 (or C0\mathcal{C}^0). It is the most basic regularity class — functions "nice enough" for optimization at least need to be continuous.

Formal View

Definition 10.8 — Continuity
A function ff is continuous at $x_0$ if limxx0f(x)=f(x0)\lim_{x \to x_0} f(x) = f(x_0). It is continuous on a set $D$ if it is continuous at every point of DD. The class of continuous functions on DD is denoted C0(D)C^0(D) (or simply C0C^0).

Polynomials, sin\sin, cos\cos, exe^x, ln(x)\ln(x) (for x>0x>0) are all continuous on their natural domains.

Interactive Visualization

Secant → Tangent Line

Why This Matters

Continuity is the minimum regularity needed for optimization — discontinuous functions can have minimizers that are impossible to find by descent methods.

  • Extreme Value Theorem: a continuous function on a closed bounded interval always attains its minimum and maximum.
  • Neural network activations must be continuous to allow gradient-based training (mostly — ReLU has a single discontinuity in derivative, not the function itself).
  • Physical models: conservation laws ensure physical quantities are continuous in space and time.

Quiz

Question 1

Which condition is NOT required for ff to be continuous at x0x_0?

Question 2

Every polynomial is a continuous function.

Common Mistakes

  • Thinking differentiability implies continuity — it does! But continuity does NOT imply differentiability.
  • Confusing a removable discontinuity (limit exists, value is wrong) with a jump discontinuity (one-sided limits differ).