Base120 · Recursion

Gradient Descent Heuristic

RE · Recursion

Iteratively adjust toward improvement, even without perfect knowledge of optimal direction

Model code
RE13
Transformation
RE (Recursion) — Improve iteratively through feedback, reflection, and self-referential refinement.
Priority
13 of 20
Framework
HUMMBL Base120 — 120 mental models across 6 transformations

When to use

"Use when the optimal direction is unknown but local improvement is measurable. Iteratively adjust in the direction of improvement, even without global knowledge. Simple but can get stuck in local optima."

Example

"A team improves a process by trying small changes and keeping the ones that help. Each week, one change is tested. Over time, the process improves. The team does not know the global optimum, but the local search finds a good-enough solution."

Common misuse

"Getting stuck in a local optimum. Hill climbing finds a peak, but not necessarily the highest peak. Periodically make a random jump (simulated annealing) to escape local optima."

Also known as

Hill ClimbingLocal Search

Related models

RE9RE1DE12