ParallelCoordinates
Displays multivariable records as lines crossing one axis per dimension.
Component behavior
ParallelCoordinates renders one axis per dimension and one line per record. It supports comparison of multivariable records, including clusters and correlations.
Use cases
- Records with several numeric dimensions.
- Relationships and trade-offs among multiple measures.
- Comparing groups of runs or configurations.
Alternatives
RadarChart: a few records are compared on the same multivariable scale.
ScatterPlot: two dimensions are sufficient.
DataGrid: the view requires exact values and sorting.
Props
Declare the dimensions in axis order, then pass data where each record's values map dimension keys to numbers. Each axis auto-scales to its own extent, so unrelated units (learning rate, accuracy, batch size) sit side by side without normalising by hand.
| Prop | Type | Default | Description |
|---|---|---|---|
dimensions | ParallelDimension[] | — | Axes left to right: a key, a label, an optional fixed domain, and an optional value formatter. |
data | ParallelRecord[] | — | Records, each with a values map (dimension key → number), an optional label, group, and color. |
groups | { name, color }[] | — | Color per group, in fixed order. Ungrouped records fall back to the palette. |
width | number | 580 | Layout width in pixels; the chart scales to its container. |
height | number | 260 | Layout height in pixels. |
Examples
Training runs
import { ParallelCoordinates } from "@solarius/onset";
import type { ParallelDimension, ParallelRecord } from "@solarius/onset";