Scatter Plot

The Scatter Plot helps you understand relationships between experiment metrics.

Instead of looking at one metric at a time, you can compare two (or three) at once and immediately see patterns, trade-offs, and outliers.

Each point represents one execution.

Selecting Executions

Before configuring the plot, choose which executions you want to compare.

On the left side panel, you can:

• Add executions to the comparison

• Remove executions

• Search and filter runs

Only the selected executions appear in the scatter plot.

Define Your Plot

In the right panel, choose:

• X Axis - Any numeric metadata or parameter

• Y Axis - Another numeric metadata field

• Radius (optional) - A third numeric value that controls bubble size

If Radius is selected, points become bubbles. Larger values = larger circles.

Common Plot Setups

Train vs Validation Performance

• X: accuracy

• Y: val_accuracy

• Radius: epochs

Detect overfitting or see whether longer training improves validation.

Quality vs Loss

• X: val_loss

• Y: val_accuracy

Identify strong performers and weak outliers.

Hyperparameter Sensitivity

• X: learning_rate

• Y: val_accuracy

See which ranges actually improve performance.

Efficiency Trade-off

• X: training_time

• Y: val_accuracy

• Radius: model_size

Understand whether larger or slower models are worth it.

Why Use Scatter Plot?

Use it when you want to:

• Detect correlations between metrics

• Spot clusters of similar runs

• Identify outliers

• Understand trade-offs (quality vs speed, size vs accuracy)

Especially useful for:

• Hyperparameter sweeps

• Architecture comparisons

• Benchmarking experiments

How To Use It

1. Select executions in the left panel.

2. Open Metadata view.

3. Click Scatter Plot.

4. Choose X, Y (and optional Radius).

5. Apply filters if needed.

The plot updates instantly.