CODEWITHSUNDEEP
algorithmoptimizationmethodsnumerical-methods
Marlon Brando
Marlon Brando

Reputation: 171

Numerical vs. algorithmic methods

In the book I am currently reading, there is a distinction made between deterministic numerical methods (e.g. gradient-based: Newton's method, gradient-free: Nelder-Mead) and algorithmic methods (e.g. exact methods: exhaustive search, heuristic methods: greedy search) for optimization.

I cannot really understand the difference that has been made. If you are given a new method and want to classify it as numerical or algorithmic method, what are you looking for in order to categorize it?

Earlier in the literature it was roughly defined that an algorithm = step-by-step instructions to solve a problem. Is that not true for every numerical method as well? Aren't they in the end algorithms too? Maybe the rough definition is missing some important point here...

I would be interested in how computer scientists think about it and how the distinction is made.

Upvotes: 0

Views: 175

Answers (1)

JimOfAllTrades
JimOfAllTrades

Reputation: 65

Numerical methods are definitely also a category of algorithms. Even stochastic methods like Monte Carlo approximations using just a bunch of random samples are in the end algorithms.

To my knowledge the conventional distinction is rather between numerical and analytical approaches, see for example also in this question: https://math.stackexchange.com/questions/935405/what-s-the-difference-between-analytical-and-numerical-approaches-to-problems.

Upvotes: 0

Related Questions