Richard Bellman

Richard Bellman (1920–1984) was an American mathematician who turned a simple, almost obvious idea — that a good plan is made of good sub-plans — into one of the most powerful tools in all of computing. He also had a genius for naming things, coining both "dynamic programming" and the wonderfully ominous "curse of dimensionality."

The lasting imprint

Bellman's insight, the principle of optimality, says that whatever you do first, the rest of your choices must be optimal for the situation you land in. Written as an equation, this becomes the Bellman equation, the recursive backbone of modern control and reinforcement learning:

V(x) = \max_{a}\; \bigl[\, r(x, a) + \gamma\, V(x')\,\bigr].

The same idea powers the Bellman–Ford algorithm for finding shortest paths through a network, even when some edges have negative weight.

Bellman later admitted the name was chosen partly as camouflage. He was working at the RAND Corporation under a Secretary of Defense who, by Bellman's telling, "had a pathological fear of the word research." So he picked "dynamic" — which sounds active and impressive and is impossible to use in a pejorative sense — and "programming," and hoped nobody in Congress would object to the budget. The math was serious; the name was a little bit of bureaucratic sleight of hand.