University of Washington
Simplex, interior-point, and memoryless quasi-Newton (QN) optimization algorithms are each viewed from two contrasting perspectives: the first facilitates computer implementation but runs counter to intuition, the second is both insightful and efficiency-revealing. Our focus will be on the memoryless QN case where the discussion is illustrated by some basic numerical experiments. Implications for limited-memory QN algorithms are briefly considered.
University of Washington