SDLabs provides several visualizations to help you track progress, explore the design space, and understand relationships between parameters. Charts are spread across the Overview and Data tabs of your experiment.
Progress Prediction Chart
Tab: Overview
Shows observed measurement values alongside the model's predicted values over iterations, with uncertainty bands. A best-seen line highlights the best observed value so far, making it easy to track whether recent iterations are improving on the current best. Use this chart to see whether the optimization is converging toward better results and how well the model's predictions match reality.
Use the objective selector to switch between objectives when your experiment has more than one.
Available: once data has been collected (at least one iteration with measurements).
Landscape Chart
Tab: Overview
A 2D projection of the design space that shows where the optimizer has sampled, what it has measured, and where it plans to look next. Points are color-coded by objective value or uncertainty, depending on the selected visualization mode.
Two visualization modes are available via a toggle:
Prediction — colors represent the model's predicted objective values across the design space.
Uncertainty — colors represent the model's uncertainty, highlighting under-explored regions.
Click Generate Landscape to create the chart for the first time. If the data changes after the chart was generated, click Update Chart to refresh it.
Available: once enough data has been acquired for the model to be built (after the initial exploration phase). Not available for experiments with constraints.
Parallel Coordinates Chart
Tab: Data
A multi-dimensional view of all parameters and measurements in your experiment. Each vertical axis represents a variable or measurement, and each line represents a datapoint. This chart is useful for spotting patterns, correlations, and outliers across many dimensions at once.
Use the objective selector to color lines by a specific objective's value. Click Toggle columns to show or hide individual axes. You can brush (click and drag) on any axis to filter the data — the selection is synchronized with the data table below.
Available: once any data has been collected.
Prediction Explanation Chart
Tab: Overview (inside Predictions section)
A waterfall chart showing how each parameter contributes to a specific predicted value. Each bar represents one parameter's contribution in shifting the prediction from the dataset average to the final predicted value.
This chart is generated on demand — query a prediction first, then click the chart icon on that row. See Insights & Predictions for details on using the prediction tool.
Available: after querying a prediction, for experiments with at most 9 parameters and 6 objectives.
When Each Chart Becomes Available
Chart | Requires |
Progress Prediction | At least one iteration with measurements |
Landscape | Predictive model (after initial exploration); no constraints |
Parallel Coordinates | Any collected data |
Prediction Explanation | Predictive model + a queried prediction; ≤9 parameters, ≤6 objectives |
Good to Know
The Landscape chart uses a dimensionality-reduction technique to project the full parameter space onto two dimensions. Points that are close together in the chart have similar parameter values.
Parallel Coordinates selections are linked to the data table — filtering in the chart filters the table, and vice versa.
All charts update automatically when new data is submitted, except the Landscape chart which must be refreshed manually.