Optimization should begin with a question you can explain and end with a result you can defend.
Between those points lies a practical craft: choosing a representation, exploiting structure, inspecting
an update and measuring what the computation actually did. This cookbook teaches that craft
through short methods you can read beside their mathematics.
You do not need to memorize 150 names. You need to recognize when a derivative, a graph, a
constraint, a population or a probability model gives you useful structure. The recipes give you
concrete places to start and equally concrete reasons to stop, debug or choose another method.
A successful exit flag does not repair a violated assumption. Keep the best value found, the
evidence of optimality and the evidence that your model fits the application as three separate claims.
That habit is more valuable than any one optimizer.