onlinerake

Real-time survey weighting for streaming data.

You're collecting survey responses or observational data one record at a time. Your sample doesn't match population demographics—too many young respondents, too few from certain regions. Traditional weighting methods (raking/IPF) require reprocessing the entire dataset whenever a new response arrives.

onlinerake updates weights as observations arrive, reducing squared error between weighted margins and population targets. Each update revisits all retained observations: memory is O(n × d), work per arrival is O(n × d × n_sgd_steps), and total work grows quadratically with stream length for fixed feature count. Check the remaining margin errors; a finite number of updates need not balance all targets.

• Online surveys where responses arrive continuously
• A/B tests that need demographic balance during collection
• Passive data collection (app usage, sensor data) requiring real-time calibration
• Moderate streams where updating retained weights after every arrival is affordable

pip install onlinerake

from onlinerake import OnlineRakingSGD, Targets

# Define population targets (proportion with indicator = 1)
targets = Targets(
    female=0.51,  # 51% female in population
    college=0.32,  # 32% college educated
    age_65_plus=0.17,  # 17% age 65+
)

# Create raker
raker = OnlineRakingSGD(targets, learning_rate=5.0)

survey_stream = [
    {"female": 1, "college": 0, "age_65_plus": 0},
    {"female": 0, "college": 1, "age_65_plus": 1},
    {"female": 1, "college": 1, "age_65_plus": 0},
    {"female": 0, "college": 0, "age_65_plus": 0},
]

# Replace this list with your incoming responses.
for response in survey_stream:
    raker.partial_fit(response)

    # Check current state anytime
    print(f"Weighted margins: {raker.margins}")
    print(f"Effective sample size: {raker.effective_sample_size:.0f}")

# Get final weights
weights = raker.weights

Both methods accept records one at a time and initialize each new weight at one. Unequal starting weights are not supported. Each method takes a fixed number of steps toward the target means, so always inspect the remaining error. The multiplicative method need not choose the same weights as a method that fits the whole dataset at once. BatchIPF provides that comparison for binary targets.

Learning rates need evaluation on the intended data, particularly when continuous features use different units. Better target balance can require more unequal weights, reducing effective sample size.

Run python examples/streaming_mwu.py for a reproducible evaluation with 10 seeds and 500 arrivals per stream. With the default MWU settings, mean squared margin error over arrivals 100–500 falls by 83% in the stationary binary case, 75% after a sampling shift, and 83% with binary and scaled continuous features. Mean final effective sample sizes are 292, 383, and 155 out of 500, respectively. These simulations measure calibration on the accumulated sample; they do not establish improved outcome estimates or a general performance guarantee.

Target means instead of proportions:

continuous_targets = Targets(
    age=(42.0, "mean"),  # Target mean age = 42
    income=(55000, "mean"),  # Target mean income = $55,000
    female=0.51,  # Binary: 51% female
)

To inspect learning-rate summability (not a guarantee of calibration):

from onlinerake import OnlineRakingSGD, Targets, PolynomialDecayLR
from onlinerake.convergence import verify_robbins_monro

schedule = PolynomialDecayLR(initial_lr=10.0, power=0.6)
scheduled_raker = OnlineRakingSGD(targets, learning_rate=schedule)

# Verify Robbins-Monro conditions (analytical for known schedules)
result = verify_robbins_monro(schedule)
print(result.condition_1_satisfied)  # True: Σ η_t = ∞
print(result.condition_2_satisfied)  # True: Σ η_t² < ∞

verify_robbins_monro() checks the two learning-rate series for known schedule types. A positive min_lr floor fails square summability. These checks do not establish convergence for a changing, clipped calibration problem.

.converged reports whether current squared moment loss is at most 1e-6 after convergence_window observations; later data can reset it. For a different loss threshold, call .check_convergence(tolerance=...). Batch IPF instead checks the maximum absolute margin error against its tolerance. A stalled or infeasible fit does not count as converged.

Unsupported numerical regret/loss-bound helpers have been removed. The learning rate helper is now suggest_mwu_learning_rate, explicitly an empirical tuning heuristic.

from onlinerake import check_target_feasibility, compute_design_effect

# Screen for missing support; this does not prove joint feasibility
feasibility = check_target_feasibility(raker)
print(f"Passes support screen: {feasibility.is_feasible}")

# Measure weighting efficiency
deff = compute_design_effect(raker)
print(f"Design effect: {deff:.2f}")

Compare streaming results against traditional IPF:

from onlinerake import BatchIPF

batch_raker = BatchIPF(targets)
batch_raker.fit(survey_stream)

print(f"Online loss: {raker.loss:.6f}")
print(f"Batch loss: {batch_raker.loss:.6f}")

Targets(**features) - Define population margins

• Binary features: female=0.51 (proportion = 1)
• Continuous features: age=(42.0, "mean") (target mean)

OnlineRakingSGD(targets, learning_rate=5.0) - SGD-based streaming raker

• .partial_fit(obs) - Process one observation
• .margins - Current weighted margins (dict)
• .loss - Current squared-error loss
• .weights - Copy of the active weight array
• .effective_sample_size - ESS accounting for weight variation
• .converged - Whether loss is below tolerance

OnlineRakingMWU(targets, learning_rate=1.0) - Multiplicative weights raker

• Same API as OnlineRakingSGD

See the 3.0.0 upgrade notes for removed bound helpers, the renamed learning-rate heuristic, changed decay floors, and corrected convergence reporting.

pip install onlinerake

Development install:

git clone https://github.com/finite-sample/onlinerake.git
cd onlinerake
uv sync --all-groups

pytest tests/ -v

See examples/ for complete worked examples:

• streaming_mwu.py - Online multiplicative weighting, drift, and the ESS tradeoff
• real_survey_example.py - Basic survey weighting
• ab_test_calibration.py - Balancing treatment/control groups
• ad_targeting_calibration.py - Real-time ad delivery calibration
• recommendation_balancing.py - Content recommendation fairness

Interactive notebooks in docs/notebooks/:

• 01_getting_started.ipynb - Visual introduction
• 02_performance_comparison.ipynb - Algorithm benchmarking
• 03_advanced_diagnostics.ipynb - Convergence and diagnostics

If you use this package in research, please cite:

@software{onlinerake,
  author = {Sood, Gaurav},
  title = {onlinerake: Streaming Survey Raking},
  url = {https://github.com/finite-sample/onlinerake},
  year = {2026}
}

MIT

See the software audit for reproduced defects, regression checks, and the relationship to the retired adaptive-eb prototype. The historical manuscript in ms/ describes an update that changes only the newest row and reports O(Kp) cost; it is not documentation of this implementation and its performance and convergence claims should not be used for this package.