Optimization History
Overview
The Optimization History chart shows how each objective function value changes over the sequence of trials. It helps you assess whether the optimization has converged, is still improving, or has stalled.
Controls
| Item | Description |
|---|---|
| Objective (multi-objective studies) | Use the dropdown to select which objective to display. |
| Moving Average | Toggle the moving average series to smooth out trial-to-trial variance. |
| Log Scale | Toggle log-scale on the value axis. |
Operations
Pan, zoom, reset, legend toggling, point hover, and click-for-detail follow the standard 2D chart controls.
| Item | Description |
|---|---|
| Series | The legend toggles the All Trials, Best Value, and Moving Average series; the point tooltip shows the trial number and objective value. |
How to Read
Reading the Chart Elements
- All Trials scatter: each dot represents one trial's objective value at that trial number.
- Best Value line: the cumulative best objective value — this line only moves when a new best is found.
- Moving Average line: a smoothed average over a sliding window of recent trials, useful for seeing the overall trend through noisy data.
Interpreting the Results
For minimization problems, a downward trend means the optimization is still progressing. If the values stop decreasing and remain flat, the optimization has converged.
When the plot appears converged, this only shows convergence under the current settings. To judge whether the solution is globally optimal, compare runs with multiple settings or use other widgets for additional analysis.
In multi-objective optimization, each objective function is shown as a separate series in this chart. When objectives trade off against each other, the plot may continue to show variation instead of a clean convergence pattern. In that case, use widgets such as Pareto Scatter and Parallel Coordinate to analyze the relationships among objectives.