Optimization problems may contain several points where the objective function is locally smallest or largest. A local minimum is a point whose objective value is no greater than the values at sufficiently nearby feasible points. A global minimum is a point whose objective value is no greater than the value at every feasible point in the entire domain. The corresponding definitions apply to local and global maxima. A global optimum is also a local optimum under the usual definition, but a local optimum need not be global. Nonlinear and non-convex optimization problems may contain many local optima, making it difficult for algorithms to find the global best solution. Convex optimization has especially useful properties: for a convex objective in a minimization problem over a convex feasible set, every local minimum is global.
Local and Global Optima
A local optimum is best within a nearby region, whereas a global optimum is best over the entire feasible domain.
Local minimum: f(x*) ≤ f(x) for all feasible x sufficiently near x*
Global minimum: f(x*) ≤ f(x) for all x ∈ S
Local maximum and global maximum use the reversed inequalities.
Explanation
Variables
x*: candidate optimum
f(x): objective function
S: feasible set
Neighborhood: a region surrounding a point
Example or application
A hilly landscape may contain several valleys. The lowest point in one valley is a local minimum, while the lowest point across the entire landscape is the global minimum.
Definition
A local optimum is a feasible point that is optimal within some neighborhood of that point, while a global optimum is optimal over the entire feasible set.
Purpose
The purpose is to distinguish nearby optimality from overall optimality and understand why some algorithms may stop at a solution that is not globally best.
Importance
This distinction is essential in nonlinear optimization, machine learning, engineering design, and other problems where objective functions may have multiple peaks or valleys.
Learning Objectives
Define local and global optima.
Distinguish local optimality from global optimality.
Recognize why non-convex problems may have multiple local optima.
Explain the importance of convexity.
Interpret local and global solutions in graphical examples.
Key Concepts
Local minimum
Global minimum
Local maximum
Global maximum
Neighborhood
Convexity
Non-convex optimization
Working Principle
Local optimality compares a point with nearby feasible points, whereas global optimality compares it with every feasible point. Mathematical properties such as convexity can establish when local optimality guarantees global optimality.
Process
Define the objective function and feasible domain.
Identify candidate local extrema.
Examine the objective values in neighborhoods around each candidate.
Compare candidate values across the full feasible domain.
Use convexity or other mathematical conditions when available.
Determine whether the best candidate is local only or globally optimal.
Applications
Nonlinear engineering design
Machine learning loss minimization
Portfolio optimization
Chemical process optimization
Scientific parameter estimation
Worked Example
Consider f(x) = x⁴ − 2x². Its derivative is f'(x) = 4x³ − 4x = 4x(x² − 1). The stationary points are x = −1, 0, and 1. The function has local minima at x = −1 and x = 1, both with value −1, and a local maximum at x = 0 with value 0. Since the function tends to positive infinity as |x| increases, the two minima are also global minima.
Common Misconceptions
Every local optimum is not necessarily global.
A global optimum can occur at a boundary rather than at a stationary point.
Non-convex problems may have multiple local optima.
A local optimum does not automatically indicate that an algorithm has failed.
Practice Questions
Define a local minimum and a global minimum.
Can a function have several local minima but only one global minimum?
Why is global optimization difficult for some non-convex functions?
What relationship does convexity establish between local and global minima?
Find the stationary points of f(x) = x³ − 3x and classify them.
Key Takeaway
Local optimality concerns nearby decisions, while global optimality concerns the entire feasible domain.
Prerequisites
Functions and graphs
Differentiation
Optimal Solutions
Reference
Boyd, S., and Vandenberghe, L. Convex Optimization.
Nocedal, J., and Wright, S. J. Numerical Optimization.
Related Topics
Optimal Solutions
Maximization and Minimization Problems
Constrained and Unconstrained Optimization
Optimization Algorithms and Solution Methods
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.
Suggest a correction or source · More in Optimization & Operations Research