Active Learning Sample Budget Calculator

Work out active learning sample budget instantly with clear inputs, formula shown and shareable results.

About Active Learning Sample Budget Calculator

Active learning spends a labelling budget in rounds, retraining after each batch so the next query targets the most informative examples. The budget is the seed set plus batch size times rounds, and coverage of the pool is usually only a few percent — which is the entire point, since the gain comes from labelling the uncertain few rather than the redundant many.

Formula

Budget

queried = min(pool, batch x rounds); total labels = seed + queried; coverage = queried / pool

Worked example

With the default values below, Active Learning Sample Budget Calculator returns Labels consumed by the whole campaign of 6,000. Change any input above and every figure here is recalculated in your browser.

Inputs used

Unlabelled pool sizeminimum 1: 100,000
Items queried per roundminimum 1: 500
Rounds plannedminimum 1: 10
Seed labels before round 1minimum 1: 1,000

Results

Labels consumed by the whole campaign: 6,000
Share of the pool labelled: 5.000%
Rounds needed to label the entire pool: 200
Labels per round: 500

Step by step

1Values used
2Budget
3Labels consumed by the whole campaign= 6,000
4Share of the pool labelled= 5.000
5Rounds needed to label the entire pool= 200
6Labels per round= 500

Assumptions and limits

The result is derived only from the 4 inputs listed above. Anything not among them is not part of the calculation.
In particular, this calculator does not take inflation, tax, fees or charges or compounding frequency into account.
Figures are rounded for display to 3 decimal places, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.

How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.

Frequently Asked Questions

How large should the query batch be?

Large enough that retraining cost is amortised, small enough that the model actually improves between rounds. One to two percent of the current training set per round is a common starting point.

Why is a seed set needed?

Uncertainty sampling requires a model. Without a few hundred to a few thousand random labels first, the initial model is too poor to rank informativeness usefully.