Scatter 2D

Sample Two Functions

Ten points are enough to see the difference between the sine and cosine curves. Each scatter() call adds a series, with its own color and legend label.

import klyn.math
import klyn.gui.windows.charts

x = NDArray<Double>.linspace(-Math.PI, Math.PI, 10)
chart = KlynPlot("2D scatter")
chart.title = "Samples of sine and cosine"
chart.xLabel = "Angle (rad)"
chart.yLabel = "Amplitude"
chart.scatter(x, x.sin(), label="Sine")
chart.scatter(x, x.cos(), label="Cosine")
chart.legend()
chart.grid()
chart.show()

Coordinates and Marker Size

scatter(x, y, label="", color="", size=36.0) accepts two one-dimensional NDArray<Double> arrays or two IList<Double> lists of equal length. The chart copies the coordinates. NaN and infinite pairs are omitted. Invalid lengths and non-finite or non-positive marker sizes raise ValueException.

chart.scatter([1.0, 2.0, 3.0], [2.0, 4.0, 3.0],
              label="Observations", color="#2ca02c", size=64.0)
chart.xLim(0.0, 4.0)
chart.yLim(0.0, 5.0)
chart.legend("upper left")

size is an area in square points, not a radius in pixels. The renderer uses a 100 DPI reference: doubling the marker diameter means multiplying size by four. The optional color uses the syntax of Klyn's Color class, such as hexadecimal or rgba(...).

Curves and scatter series can share the same 2D chart. Each kind maintains its own color cycle. grid(false) hides the grid; autoscale() restores automatic limits, and clear() removes all series.

Scalar Colors, Areas and Colorbars

chart = KlynPlotWidget()
chart.scatter(longitudes, latitudes, label="House values",
              values=houseValues, sizes=populationAreas,
              colorMap="coolwarm", vmin=0.0, vmax=5.0,
              sortByValue=true)
chart.colorbar("Median house value (100,000 USD)")
chart.legend()

values and sizes are optional IList<Double> lists with one entry per coordinate pair. Scalar values must be finite. Areas must be finite and non-negative; zero hides a marker. Scalar coloring cannot be combined with a fixed color. All inputs are copied when adding the series.

The coolwarm palette goes from blue through a light neutral midpoint to red. Limits default to the scalar minimum and maximum; supplied vmin and vmax must be finite and non-decreasing. Values outside them are clamped. Equal limits use the middle color. sortByValue=true draws low values first without separating colors, sizes and coordinates; equal values keep their original order.

Colors and optional sorted indices are cached once on insertion, not on repaint. colorbar(label) uses the most recently added scalar-colored series' exact palette and limits. Use equal limits on related charts for a meaningful comparison. colorbar(visible=false) hides it; clear() clears it along with the plotted data.

To compare observations with predictions, choose shared color limits from the observed targets and reuse them unchanged on both charts. Including prediction outliers in these limits would change the colors of the observed data too. Out-of-range predictions saturate at the palette endpoints; their numeric values and the model's loss are not clipped. Blue and red represent value ranges, not equal-sized groups of observations.

In samples/math/nn/CaliforniaDNN.kn, only the model's input features are standardized. Geographic coordinates and population areas are copied before in-place scaling; observed targets and predictions both use units of 100,000 USD. The chart never receives standardized geography or a mixture of target units.

Use It in Your GUI

Replace KlynPlot() with KlynPlotWidget() and add it to a window layout. The drawing API and visual result are the same; only the parent window runs the event loop. See the embedding example.

With a native GUI window open, chart.exportImage(Path("scatter.png")) exports a 640x480 image. Import klyn.io for Path.

klyn samples/gui/charts/Scatter2DSample.kn

Add a third coordinate to each observation in Scatter 3D.