Getting Started with SigOpt

Welcome to SigOpt. In this tutorial you will connect to SigOpt’s optimization service and use it to find the maximum of an example function.

SigOpt accelerates machine learning model tuning and industrial process engineering, reducing R&D time and resources invested in model tuning or process optimization by up to 100x. We use Franke's function, a two-dimensional optimization problem, to demonstrate SigOpt’s optimization feedback loop.

We first create a two-dimensional Experiment in SigOpt with two double Parameter objects, x and y. Note that we don’t pass any other information about the function to SigOpt.

Then, we create a feedback loop to find the maximum of the function using SigOpt.

  1. Receive a Suggestion from SigOpt’s API.
  2. Evaluate the function using the parameters from the Suggestion.
  3. Report an Observation back to SigOpt.
Step 0: Create an Experiment
First, we'll create a SigOpt Experiment with two tunable parameters, x and y.
NameFranke Optimization
Parameters
NameTypeRange
xDecimal[0,1]
yDecimal[0,1]
Step 1: Receive a Suggestion
To start the feedback loop, we’ll ask SigOpt’s API for a Suggestion. SigOpt will send back values for x and y.
Parameter x:
Parameter y:
Step 2: Evaluate Your Metric
The metric is a representation of the value that the Experiment is optimizing. To evaluate the metric, we plug the assignment values from our Suggestion into the Franke function.
Franke Function
Value:
Step 3: Report an Observation
Now, we’ll submit an Observation with the metric value we computed in the last step. Typically you’d report the accuracy/AUC of a machine learning model or the measured output of a physical process.
Step 4: View Results
Now, view the progress, results, and history of your experiment on SigOpt’s live dashboard.
Step 5: Developer Documentation

Learn more with the developer documentation for our API Client.