Sample Selection

Overview

Use Generate > Samples Selection… to generate multiple representative composite subsets from one composite set.

Each generated selection is saved as a subset of the source composite set (not a full copy), so it is lighter and faster to work with. In the composite list, subsets are shown in green.

This tool is available for composite sets that support the sample-selection workflow.

Before You Start

• Select exactly one composite set.
• If you plan to use weighted classes, run declusterization first so weights are available.
• If you plan to filter subsets later, generate global special variables first (for example averages and standard deviations).

General Settings

• Folder: Destination folder for generated selections.
• Number of selections: Number of selections to generate.
• Use only a subset: Limits candidate composites to a specific tag subset.
• Subset tag: Tag used when Use only a subset is enabled.
• Add a forced subset: Forces all composites of one tag to be included in every selection.
• Forced tag: Tag used when Add a forced subset is enabled.
• Define selections size: Strategy used to target a selection size.
• Selections size: Target number of composites per selection (used when size strategy is enabled).
• Number of spatial classes: Number of spatial blocks to enforce as coverage classes. 0 means no spatial classes.

Size Strategies

When Define selections size is enabled, one of the following strategies is used:

• None: No explicit size targeting.
• Trick: Increases variable class constraints to move generated selection sizes toward the target.
• Combine: Combines generated smaller selections to build selections closer to the requested size.
• Force: Keeps only selections that exactly match the requested size.
• Naive: Uses a combinational search approach up to the requested size; can be slower on large datasets.

Constraint Types Used by the Selector

A generated selection must satisfy all active classes from the enabled constraint types.

1) Variable Classes

For each selected variable:

• <Variable> classes defines the number of histogram bins (classes).
• A valid selection must include at least one composite in each active class.
• If Use only a subset is enabled, empty classes (with no candidate composite) are automatically ignored.
• Setting classes to 0 disables that variable in selection-class creation.

Optional:

• <Variable> use weights builds classes using decluster weights (equal weight mass per class rather than equal composite count).

2) Tag Classes

For any variable marked as a tag class:

• each distinct integer tag value becomes one class,
• each selection must include at least one composite for every retained tag class.

3) Envelope Classes

You can define up to 20 envelope-set entries.

• Each entry accepts a comma-separated list of envelope names.
• Each envelope in that list creates one class.
• A valid selection must include at least one composite inside each active envelope class.

4) Spatial Classes

If Number of spatial classes is greater than 0:

• the 3D space is partitioned into spatial classes,
• each selection must cover all retained spatial classes,
• empty classes are ignored automatically,
• isolated composites can be rejected during spatial class construction.

How Results Are Generated

• The selector generates as many unique valid selections as possible up to the requested count.
• If fewer selections can be generated than requested, Genesis reports this and lets you decide whether to keep the available selections.
• Logs are written in each generated subset, including achieved constraints and a theoretical minimum reference size.

Special Variables for Composite Sets and Subsets

Use Generate > Special Variables… to add extra variables.

Spatial Variables

• Block Coverage Tag (BCT): Ordinary (per-composite) variable indicating the block index covered by each composite.
• Space Coverage Index (SCI): Global variable indicating spatial dispersion of a set/subset.

Statistical Global Variables

• Average: Adds avg_<variable> global variables.
• Standard Deviation: Adds dev_<variable> global variables.
• Percentiles: Adds per_<p>_<variable> global variables (multiple percentile levels).

A weighting mode can be used for average/standard deviation (for example Declust Weight or another variable).

Tag and Binary Variables

• Tag variables (for example tag_litho, tag_alter, tag_miner, or other dictionaries) add integer class identifiers per composite.
• Binary variables create one 0/1 variable per dictionary entry.

Composite Subsets

Subsets can be created by sample selection or with Subsets > Create Subset.

You can then:

• edit subset membership with Subsets > Edit Subset…,
• combine subsets with Subsets > Union…,
• keep only common elements with Subsets > Intersection…,
• remove elements with Subsets > Substraction….

Filtering Subsets with Global Variables

Use Subsets > Filters… to keep only subsets within value ranges of global variables.

Typical workflow:

1. Generate special global variables first (for example avg_*, dev_*, percentiles).
2. Open Subsets > Filters….
3. Set:
   • Output Folder,
   • per-variable enable/disable,
   • Minimal Value and Maximal Value thresholds.
4. Keep only subsets that satisfy all enabled variable ranges.

This filtering compares each subset’s global values to your accepted ranges (often based on the population reference values shown in the filter dialog).