Case Studies

End-to-end active learning campaigns that combine the building blocks of AFL-agent into complete workflows on virtual instruments.

Bioformulations

Phase mapping of a protein / glycerol / temperature formulation on the ProteinGlycerolTemperaturePhaseMap virtual SAS instrument. Both notebooks run a single campaign with the same Dirichlet GP phase classifier and the same F1 benchmark against the simulator; they differ in how the next composition and temperatures are chosen. The third notebook runs the manifold campaign through AFL servers: the virtual instrument and the agent run as AFL-automation APIServers that exchange measurements through Tiled.

• Bayesian Active Learning of a Protein Formulation Phase Map
• Manifold Active Learning of a Protein Formulation Phase Map
• Closed-Loop Manifold Active Learning with AFL Servers