- Overview
- Summary
- Funding
- Cite This Publication
Overview
Who: Zachary Cosenza & David E. Block
Published: October 27, 2020
Where: Engineering Optimization
Key Takeaway: A new machine learning algorithm could reduce the time and resources needed for complex experiments. The approach could be applied to optimizing cell culture media.
Research Topics:
Summary
Zachary Cosenza & David E. Block discuss the development of a computational model that is capable of optimizing design parameters for physical and biological processes. They test two methods, a neural network genetic algorithm (NNGA) and a dynamic coordinate search for response surface methods (DYCORS), against a hybrid model combining the two. Cosenza & Block find that the hybrid NNGA-DYCORS learning algorithm outperforms the individual models. Their findings also hold up with the addition of noise to the data, a more realistic test of the model’s use. They argue the hybrid model is robust and generalizable, and could be usable for many types of design problems, regardless of familiarity with surrogate modeling.
Written by Elie Diaz
Funding
Cite This Publication
Cosenza, Z., & Block, D. E. (2020). A generalizable hybrid search framework for optimizing expensive design problems using surrogate models. Engineering Optimization, 1-14. doi:10.1080/0305215x.2020.1826466
Copied!