Results¶

These reference results are generated from the public package APIs used by the notebooks. The notebooks remain thin clients; the API path is used here because it is deterministic, CI-friendly, and avoids committing executed notebook outputs.

The configurations are intentionally small so the GitHub Pages workflow can refresh the page quickly. They are reproducible smoke-scale examples, not quantum-advantage claims.

Environment¶

Generated: 2026-05-06 11:07:15 UTC
Git commit: 9496565
Python: 3.12.13
Package version: 0.1.12
PennyLane: 0.44.1
Matplotlib backend: Agg
Default execution: analytic default.qubit unless a shot count is listed

Summary¶

Workflow	Primary metric	Value	Runtime
Variational quantum classifier	`train_accuracy`	0.4595	19.14 s
Variational quantum regression	`train_mse`	0.9098	2.19 s
Quantum convolutional neural network	`train_accuracy`	0.7000	77.74 s
Quantum autoencoder	`test_compression_fidelity`	0.7014	2.09 s
Quantum kernel classifier	`train_accuracy`	0.8519	2.18 s
Trainable quantum kernel	`train_accuracy`	0.7333	17.19 s
Quantum metric learning	`train_accuracy`	0.5946	2.52 s

Variational quantum classifier¶

Configuration:

dataset=moons, n_samples=50, noise=0.1000, seed=123, n_layers=1, steps=8, shots=analytic

Metric	Value
`train_accuracy`	0.4595
`test_accuracy`	0.6154
`final_loss`	1.4790
`runtime_seconds`	19.14

Images:

moons embedangle layers1 steps8 samples50 noise0p1 seed123 analytic dataset

Variational quantum regression¶

Configuration:

dataset=linear, n_samples=50, noise=0.1000, seed=123, n_layers=1, steps=8, shots=analytic

Metric	Value
`train_mse`	0.9098
`test_mse`	0.3316
`final_loss`	0.9841
`runtime_seconds`	2.19

Images:

linear layers1 steps8 samples50 noise0p1 seed123 analytic dataset

Quantum convolutional neural network¶

Configuration:

dataset=moons, n_samples=40, noise=0.1000, seed=123, steps=6, shots=analytic

Metric	Value
`train_accuracy`	0.7000
`test_accuracy`	0.9000
`final_loss`	0.5754
`runtime_seconds`	77.74

Images:

moons steps6 samples40 noise0p1 seed123 analytic dataset

Quantum autoencoder¶

Configuration:

family=correlated, n_samples=32, noise=0.0500, seed=123, n_layers=1, latent_qubits=2, steps=6

Metric	Value
`test_compression_fidelity`	0.7014
`test_reconstruction_fidelity`	1.0000
`final_loss`	0.3676
`runtime_seconds`	2.09

Images:

correlated layers1 latent2 steps6 samples32 noise0p05 seed123 loss

Quantum kernel classifier¶

Configuration:

dataset=moons, n_samples=36, noise=0.1000, seed=123, shots=analytic

Metric	Value
`train_accuracy`	0.8519
`test_accuracy`	0.8889
`runtime_seconds`	2.18

Images:

moons samples36 noise0p1 seed123 analytic dataset

Trainable quantum kernel¶

Configuration:

dataset=moons, n_samples=20, noise=0.1000, seed=123, embedding_layers=1, steps=2, shots_train=analytic, shots_kernel=analytic

Metric	Value
`train_accuracy`	0.7333
`test_accuracy`	0.8000
`final_alignment`	0.1755
`final_loss`	-0.1755
`runtime_seconds`	17.19

Images:

moons trainable kernel embdata reupload layers1 steps2 samples20 noise0p1 seed123 analytic analytic alignment

Quantum metric learning¶

Configuration:

dataset=moons, samples=50, seed=42, layers=1, steps=8, pairs_per_step=16, log_every=0

Metric	Value
`train_accuracy`	0.5946
`test_accuracy`	0.6923
`final_loss`	0.1152
`runtime_seconds`	2.52

Images:

moons layers1 steps8 samples50 margin0p5 seed42 embeddings

Reproduce¶

Regenerate this file from the repository root:

python docs/pages/generate_results.py

The GitHub Pages workflow also regenerates this file before building the web pages. Generated images are written under docs/pages/assets/reference-results/ and embedded above.