Welcome to Atlas’ documentation!

Atlas is an chemistry and materials science adaptive experimentation package.

from olympus import Surface, Campaign
from atlas.planners.gp.planner import GPPlanner

surface = Surface(kind='Branin') # instantiate 2d Branin-Hoo objective function

campaign = Campaign() # define Olympus campaign object
campaign.set_param_space(surface.param_space)

planner = GPPlanner(goal='minimize', num_init_design=5) # instantiate Atlas planner
planner.set_param_space(surface.param_space)

while len(campaign.observations.get_values()) < 30:
    samples = planner.recommend(campaign.observations) # ask planner for batch of parameters
    for sample in samples:
        measurement = surface.run(sample) # measure Branin-Hoo function
        campaign.add_observation(sample, measurement) # tell planner about most recent observation

Getting Started

• Overview
• Installation
• Credits
• Support

Examples

• Single objective optimization
• Optimization with a priori unknown constraints
• Multi-objective optimization with achievement scalarizing functions
• Multi-fidelity optimization

Core Classes

• Planners
• Gaussian processes
• Acquisition Functions
• Acquisition Optimizers
• Known Constraints