Random Search

Random Search samples parameter values randomly instead of testing every combination.

This is more efficient when you have many parameters or large ranges, you'll explore the space faster without exhaustive enumeration.

When to use Random Search

• Large parameter spaces: Testing all combinations would create too many executions

• Budget constraints: You have limited time or compute resources

• Exploratory tuning: You want a quick sense of promising regions before refining

Create a Random Search

Follow the same steps as Grid Search, but:

1. Select Random Search as the Task type

2. Open your project in Valohai

3. Go to the Tasks tab

4. Click Create Task

5. Select the step that contains your parameters

6. Scroll to Parameters

7. For each parameter, choose Random as the type

8. Specify:

   • Min and max values for the random range

   • Count: How many random samples to draw for this parameter

Valohai will generate random parameter combinations and queue the specified number of executions.

💡 Tip: Random Search often finds good hyperparameters faster than Grid Search when you have more than 3-4 parameters.