What is it about?
When looking for molecules with optimal properties under a set of constraints, a smart search strategy is very important. This paper introduces such a strategy which adapts as it searches the space to take greater advantage of any previously unrealized underlying principles.
Featured Image
Why is it important?
A discrete search space can be essentially randomly ordered even if there is a principled order possible. Extracting an order as the presented algorithm does reduces computational resources by a factor of 2 and improves the quality of the result compared to state of the art breadth-first search.
Perspectives
Not included in this paper is the analysis that the reordering methodology is approximately optimal for pair-wise functions, which makes it an effectively on-the-fly implicit dead-end elimination. The slow solution of the dual problem for the constraints really prevents early trapping and greatly improves the quality. In the age of increased focus on machine(-enabled) learning, the idea of determining and leveraging emergent principles will surely find increasing use in optimizations to come.
Dr Berend C Rinderspacher
US ARL
Read the Original
This page is a summary of: Smooth heuristic optimization on a complex chemical subspace, Physical Chemistry Chemical Physics, January 2015, Royal Society of Chemistry,
DOI: 10.1039/c5cp02177d.
You can read the full text:
Contributors
The following have contributed to this page
