Continuously Differentiable Functions (C¹)
A function is continuously differentiable (class ) if its derivative exists at every point AND the derivative function is itself continuous. This is a stronger requirement than merely being differentiable — the derivative must vary smoothly without jumps.
Why does it matter that the derivative be continuous? If has a jump discontinuity at some point, then locally the function has "different slopes from the left and right" — a kink. Kinks cause problems in optimization algorithms. functions are "smooth enough" for most gradient-based methods to work reliably.
Notation: denotes the class of functions with continuous first derivative on domain . means the first derivatives all exist and are continuous. means all derivatives exist and are continuous — these are called "smooth" functions.
Formal View
(smooth) functions include polynomials, , , , . is but not (derivative has a jump at 0). is but not .
Why This Matters
is the standard regularity assumption for gradient-based optimization — it ensures the gradient is well-defined and varies predictably.
- Gradient descent requires : the gradient must be defined and continuous for standard convergence guarantees.
- Finite element methods: solution regularity () determines approximation error rates.
- PDE theory: elliptic equations have solutions under smoothness assumptions on coefficients.
Quiz
Which statement correctly describes a function?
is in class .
Common Mistakes
- Thinking differentiable everywhere implies — the derivative must also be CONTINUOUS for .
- Confusing (first derivatives continuous) with having derivatives that are just bounded.