Skip to contents

Benchmark quantum-foundation KRR against classical baselines

Source: R/quantum_foundation.R
qf_krr_benchmark.Rd

Runs four regressors on the same train/test split:

  1. ranger (QRF) over the RAW covariates (our established plain-ML baseline from the v1.3 case-cerrado-end-to-end).

  2. RBF Kernel Ridge Regression over the foundation-embedding PCs.

  3. Quantum Kernel Ridge Regression over the RAW covariates (Pillar 6 original).

  4. Quantum Kernel Ridge Regression over the foundation-embedding PCs (the v2.0.0 contribution).

Usage

qf_krr_benchmark(
  embeddings,
  covariates,
  y,
  train_ix = NULL,
  test_ix = NULL,
  n_pcs = 8L,
  reps = 2L,
  lambda = 0.1
)

Arguments

embeddings

Foundation-model embedding matrix (n_obs x D).

covariates

Raw-covariate matrix or data frame (n_obs x C).

y

Numeric response.

train_ix, test_ix

Integer index vectors selecting rows for train/test. If NULL, a 70/30 random split is drawn.

n_pcs, reps, lambda

Passed to qf_krr_fit().

Value

A data frame with columns method, rmse, mae, r2, n_train, n_test.

Details

Returns the test-set RMSE, MAE and R^2 for each setup so the question "does the quantum lift on foundation embeddings beat either its quantum-only or foundation-only ancestor?" gets an empirical answer.