The idea
Real problems are defined partly by deadlines, latency, resources, incomplete data and other constraints. A theoretically imperfect approximation may solve the actual problem better.
Situation
The theoretically correct solution is expensive, slow or incompatible with real constraints.
Useful when
Engineering trade-offsdeadlinesprototypesresource constraints
Watch for
“Pragmatic” shouldn't become an excuse for ignoring requirements that genuinely determine success. Sometimes complexity really is necessary.
Where this came from
Pommerman, Competitions, Applied machine learning