Running an Optimization
Once you have defined your variables, measurements, and objectives, you can launch the optimization. This article walks through the optimization lifecycle — from the first suggestions to interpreting results — and links to detail articles on each part of the workflow.
The Optimization Loop
Every optimization in SDLabs follows the same iterative cycle:
Optimize — the algorithm analyzes your data and generates a batch of recommended experiments.
Experiment — you run the suggested experiments in your lab (or simulation).
Measure — you enter the results back into SDLabs.
Repeat — submitting results triggers a new recommendation cycle. The algorithm learns from the new data and produces better recommendations.
Each cycle is called an iteration. The more iterations you complete, the better the optimizer understands your system.
Two Phases of Optimization
The optimization progresses through two distinct phases, shown as a badge on the recommendations panel:
| Initial Exploration | Guided Optimization |
Badge | Orange — "Evaluating possibilities" | Blue — "Finding most informative conditions" |
What happens | The algorithm explores the design space broadly to gather diverse data points. Suggestions are spread across the parameter space. | The algorithm has built a predictive model and uses it to focus on the most promising regions. Suggestions become more targeted. |
When it starts | Immediately on the first iteration. | After enough data has been collected (typically a few iterations, depending on the number of parameters). |
During initial exploration, the badge shows how many iterations remain before the model kicks in. Once guided optimization begins, features like Predictions and Landscape charts become available.
What You See on Screen
State | What you see | What to do |
Recommending (shimmer text) | The algorithm is analyzing your data and generating recommendations. | Wait — this typically takes a few seconds to a few minutes. |
New recommendations | A batch of suggested parameter values is ready. | Run the experiments, then enter your measurements and submit. |
Detail Articles
Each article below covers a specific part of the optimization workflow in depth:
Suggestions & Measurements — how to work with recommendations, submit results, and add your own datapoints
Insights & Predictions — AI-generated explanations, the prediction query tool, and SHAP-based explanations
Analytics & Visualizations — Progress Prediction, Landscape, and Parallel Coordinates charts
Related
Experiment Setup — how to define variables and measurements before launching
Experiment Design Best Practices — parameter count, and objective strategy guidance
Multi-Objective Optimization — Pareto, Weighted Sum, and Hierarchy strategies
Expert Context — guiding the optimizer with domain knowledge