Experiment Options
A SigOpt Experiment is created as an object containing all information needed to execute the optimization process. To create a SigOpt Experiment, you will define:
Name
Type
Key
Description
Experiment Name
String
name
Name of the experiment.
Experiment Type
String
type
Identifier for the type of experiment.
Parameter Space
Array
parameters
List of objects where each object is defining a single parameter.
Metric Space
Array
metrics
List of objects where each object is defining a single metric.
Budget
Integer
budget
Integer defining the minimum number of SigOpt Runs in a given SigOpt Experiment.
Parallel Bandwidth
Integer
parallel_bandwidth
Integer defining the number of machines/nodes there will be running different asynchronous SigOpt Runs at any given time.
The SigOpt Optimizer provides four types of experiments: optimization experiment, all constraints experiment, grid search experiment and random search experiment. They are invoked by setting the type flag while creating an Experiment.
Search Experiment
Type Flag
SigOpt Optimization Experiment
offline
All Constraints Experiment
offline
Grid Search
grid
Random Search
random
In the next few pages we'll go through some examples on how to set up your SigOpt Experiment for each type of search above.
Copy link