Next generation Design of Experiment
Why use Sequential Learning?
Sequential Learning accelerates product development by reducing the number of experiments required to find new materials that meet performance specifications. This workflow is effective in projects with small, sparse data sets because it works through an iterative approach. It enables extrapolation to new, higher performance materials.
High-Dimensional Materials and Process Optimization Using Data-Driven Experimental Design with Well-Calibrated Uncertainty Estimates.
Julia Ling, Maxwell Hutchinson, Erin Antono, Sean Paradiso, Bryce Meredig. Integrating Materials and Manufacturing Innovation, 6(3), 207–21 (2017).
What is it?
Sequential Learning is an iterative data-driven method of exploring a parameter space.
- AI models are trained on existing materials data
- The models sift through the design space of possible candidate materials to surface new candidate materials for evaluation
- The product developer selects experiments to run from the suggested list, collects the data, and adds that data back in to the training set
- The model learns from the new data, gets smarter, and provides new experimental suggestions
How does it find new, higher performing materials?
Typical AI models are not effective at extrapolation. Our Sequential Learning workflow leverages state-of-the-art uncertainty estimates. These uncertainty estimates mean that our AI models can explore and extrapolate systematically. This approach means that our AI models can find new materials that outperform all the materials in the original data set.