micro-mass-quantum

Quantum-native labels and amplitude kernel for the micro-mass mass spectrometry dataset, produced by Sirius Quantum.

Compression Benchmark

At 162x fewer dimensions, ReLab achieves 92.2% accuracy versus 93.6% with full features. 1.4% accuracy loss qualifies under the oracle sketching criterion (Zhao et al. 2026, arXiv:2604.07639): compression ratio > 10x with comparable accuracy (≤5% loss).

	Method	Features	Accuracy
Classical baseline	SVM-linear	1300	93.6%
ReLab quantum kernel	SVM-amplitude	8 qubits	92.2%
Delta		162x compression	-1.4%

Claim: 8-qubit amplitude encoding achieves 92.2% on 10-class mass spectrometry identification at 162x compression, within 1.4% of SVM-linear on all 1300 features.

QQ Benchmark: Quantum Community Baseline

First published quantum kernel benchmark on micro-mass. Future quantum methods (QAOA, VQE, variational circuits, tensor networks) compare against this baseline.

Property	Value
Method	amplitude / 8q / SVM / oracle-sketching
Encoding	Amplitude (Hilbert-Schmidt fidelity)
Qubits	8
Compression	162x (1300 features → 8 qubits)
Kernel	K(x,x') = |⟨ψ(x)|ψ(x')⟩|²
CV protocol	StratifiedKFold(n_splits=5, shuffle=True, seed=42)
QQ score	92.2% accuracy (10-class)

Dataset

Source: micro-mass, OpenML ID 1515. 360 samples, 1300 mass spectrometry features, 10 bacterial species classes.

Column	Type	Description
features	list[float]	Raw mass spectrometry features (1300-dim, unscaled)
label	int	Integer class label (0–9)
class_name	str	Bacterial species name
quantum_label	int	Quantum-native label ∈ {-1, +1} derived from amplitude kernel geometry

Quantum label balance: +1=180 / -1=180 (perfectly balanced — amplitude kernel eigenvector split).

Usage

from datasets import load_dataset
import numpy as np

ds = load_dataset("SiriusQuantum/micro-mass-quantum", split="train")

X = np.array(ds["features"])           # (n, 1300) raw features
y = np.array(ds["label"])              # (n,) class labels 0-9
q = np.array(ds["quantum_label"])      # (n,) quantum labels {-1, +1}

# Train on quantum-labeled partition
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
X_std  = scaler.fit_transform(X)

# Quantum label as geometric prior
svm = SVC(kernel="rbf")
svm.fit(X_std, q)   # train on quantum partition signal

Reproduce

Quantum labels and kernel matrix were produced with the ReLab engine, a quantum-native data layer for physical AI.

import relab

# One call. Quantum-native labels from 1300 features compressed to 8 qubits.
labels = relab.fit(X, n_qubits=8, encoding="amplitude")

SDK not yet public. Contact Sirius Quantum for early access.

Scientific Basis

Encoding: Amplitude (Lloyd, Schuld 2020, arXiv:2001.03622)

• K(x,x') = |⟨ψ(x)|ψ(x')⟩|² (Hilbert-Schmidt fidelity kernel)
• Theoretically optimal when classes are separated in Hilbert space
• Scale-invariant: StandardScaler output feeds directly, no aliasing

Compression claim basis: Oracle Sketching (Zhao et al. 2026, arXiv:2604.07639)

• Polylogarithmic quantum machines match exponentially larger classical machines
• Criterion: compression ratio > 10x with comparable accuracy (≤5% loss)
• micro-mass: 162x compression, 1.4% accuracy loss. Both criteria satisfied.

Citation

If you use this dataset or kernel, cite ReLab:

@software{relab2026,
  title  = {ReLab: Quantum-Native Data Relabeling Engine},
  author = {Sirius Quantum Solutions LTD},
  year   = {2026},
  url    = {https://github.com/Sirius-Quantum}
}

Key references:

@article{lloyd2020quantum,
  title   = {Quantum embeddings for machine learning},
  author  = {Lloyd, Seth and Schuld, Maria and others},
  journal = {arXiv:2001.03622},
  year    = {2020}
}

@article{zhao2026sketching,
  title   = {Quantum oracle sketching},
  author  = {Zhao et al.},
  journal = {arXiv:2604.07639},
  year    = {2026}
}