Inspect a warm-start ensemble

Inspect a warm-start ensemble

Usage

# S3 method for class 'pairwiseLLM_warm_ensemble'
summary(object, ...)

# S3 method for class 'pairwiseLLM_warm_ensemble'
print(x, ...)

Arguments

object, x

A pairwiseLLM_warm_ensemble.

...

Reserved; must be empty.

Value

summary() returns component summaries and ensemble contract details. print() returns its input invisibly. Component metrics do not validate the ensemble. Between-model SD is diagnostic disagreement, not Bayesian prior SD.

See also

ensemble_warm_start_models(), predict.pairwiseLLM_warm_ensemble()

Other adaptive warm start: ensemble_warm_start_models(), extract_warm_start_features(), fit_warm_start_model(), make_warm_start_cv_plan(), make_warm_start_prior(), pairwiseLLM_warm_model, predict.pairwiseLLM_warm_ensemble(), predict.pairwiseLLM_warm_model(), prepare_warm_start_model(), register_warm_start_model(), save_warm_start_model(), summary.pairwiseLLM_warm_predictions(), warm_start_coefficients(), warm_start_feature_schema(), warm_start_python_status()

Examples

if (requireNamespace("glmnet", quietly = TRUE) &&
    requireNamespace("withr", quietly = TRUE)) {
  local({
    # Synthetic features illustrate the interface, not predictive validity.
    example_features <- function(seed) {
      withr::local_seed(seed)
      fields <- warm_start_feature_schema()$feature
      x <- as.data.frame(matrix(runif(15 * length(fields)), nrow = 15))
      names(x) <- fields
      x$n_tokens <- 11:25
      x$token_length_mean <- 2 + 10 * x$token_length_mean
      x$token_length_std <- 0.2 + x$token_length_std
      x$dale_chall_readability_score <- 5 + 20 * x$dale_chall_readability_score
      x <- data.frame(item_id = as.character(1:15), x)
      attr(x, "warm_start_schema") <- "writing_features_v1"
      x
    }
    features <- example_features(3103)
    theta <- 10 + 0.4 * features$n_tokens - 2 * features$token_length_mean
    # A small alpha grid keeps this example fast; the default has 41 values.
    model <- fit_warm_start_model(features$item_id, theta, "synthetic-a",
      features = features, alpha_grid = c(0, 1))
    features_b <- example_features(3104)
    theta_b <- 30 + features_b$n_tokens - 3 * features_b$token_length_mean
    model_b <- fit_warm_start_model(features_b$item_id, theta_b, "synthetic-b",
      features = features_b, alpha_grid = c(0, 1))
    ensemble <- ensemble_warm_start_models(assessment_a = model, assessment_b = model_b)
    summary(ensemble)
    print(ensemble)
  })
}
#> Warm-start ensemble: 2 equally weighted task models
#> Components: assessment_a, assessment_b 
#> Target: within_task_z; audit: full 
#> Between-model sample SD is diagnostic, not Bayesian prior SD.