Quantum and classical computing research
DeTection Inc. is exploring computational methods relevant to future matter characterization, biological detection and medical research. Our public research summaries describe early, reproducible benchmarks on openly available data—not deployed detection systems.
Amazon Braket ↗Eight-qubit local simulation ↓Photonic simulation ↓Classical benchmarks ↓Research documents — PDF reports
These downloadable reports summarize public-dataset and software-simulation results only.
Google Colab medical benchmark (PDF)Public dataset, 30-feature and eight-PCA-feature classical baselinesAmazon Braket eight-qubit local simulation (PDF)Software simulation, no physical quantum processorClassical versus quantum-derived comparison (PDF)Four-model accuracy, F1 and ROC AUC resultsPDF links work on the live website after the corresponding PDF files are uploaded alongside research.html.
Amazon Braket: eight-qubit hybrid quantum–classical benchmark
Research performed/reported: October 2026 · Website updated: November 2, 2026
In October 2026, DeTection Inc. reported an exploratory eight-qubit circuit experiment using the Amazon Braket local simulator and the public Breast Cancer Wisconsin (Diagnostic) dataset. The evaluation compared a quantum-derived classifier with conventional machine-learning baselines using a fixed split of 455 training examples and 114 held-out test examples.
Amazon Braket local simulator
Circuit-based computational benchmark; not a hardware QPU execution.
Held-out test examples
From a public dataset containing 569 total examples.
Reported held-out accuracy
| Model | Accuracy |
|---|---|
| PCA + logistic regression | 95.61% |
| Random forest | 91.23% |
| Gradient boosting | 92.98% |
| Eight-qubit quantum-derived classifier | 66.67% |
Interpretation: The classical models outperformed the quantum-derived classifier in this initial comparison. This is an exploratory technical baseline, not evidence of a quantum advantage. Further work would require carefully specified feature processing, repeated trials, noise-aware evaluation, independent replication and, if appropriate, separately approved hardware testing.
Visit the official Amazon Braket website ↗ · Amazon Braket documentation ↗
Amazon Braket is an AWS quantum-computing service. Mentioning it identifies the software environment used; it does not imply that Amazon or AWS endorses, sponsors or independently validates DeTection Inc.'s research.
Photonic quantum classification: MerLin simulation benchmark
A separate September 28, 2026 research benchmark implemented photonic classification workflows with Quandela's MerLin software framework and evaluated them on the same publicly available Wisconsin Diagnostic Breast Cancer dataset. This was a perfect-simulation study, not a photonic-hardware or clinical experiment.
2- and 3-photon FidelityKernel
Reported test accuracy on 40 held-out cases for each simulated configuration.
Quantum Reservoir Classifier
Reported 2-photon and 3-photon simulation results, respectively.
Dataset and methods
The benchmark used the Breast Cancer Wisconsin (Diagnostic) dataset, distributed through scikit-learn and originally sourced from the UCI Machine Learning Repository. It contains 569 samples, 30 numerical features and two classes. The reported setup used PCA to six features (approximately 89.2% variance retained), angle encoding with inputs scaled to [0, 1], and a subset of 80 training and 40 test cases.
| Element | Reported setup |
|---|---|
| Software | MerLin 0.4.1; Perceval-Quandela 1.3.0 |
| Configurations | 2 photons and 3 photons |
| Methods | FidelityKernel; Quantum Reservoir Classifier |
| Simulation | Perfect simulation (shots=0) |
| Data source | Public UCI benchmark, scikit-learn packaged copy |
Reported photonic simulation results
| Method | Photons | Test accuracy |
|---|---|---|
| MerLin FidelityKernel | 2 | 97.5% |
| MerLin FidelityKernel | 3 | 97.5% |
| Quantum Reservoir Classifier | 2 | 97.5% |
| Quantum Reservoir Classifier | 3 | 95.0% |
Classical reference and interpretation
The original photonic report cited approximately 98.60% accuracy for full 30-feature logistic regression and 97.90% for an RBF support vector machine on its fixed comparison split. The photonic simulations therefore did not establish superiority over those classical references. These figures belong to a different experimental setup from the Amazon Braket study above and should not be compared as if they used the same train/test protocol.
Research direction and limitations
These public benchmarks inform DeTection Inc.'s exploration of hybrid quantum–classical methods. Potential future studies may investigate sensing, matter characterization and biological detection, subject to independent testing and appropriate safeguards. No claims of real-world sensitivity, specificity, diagnostic accuracy, clinical utility, regulatory authorization or commercial readiness are made from these simulations.
Public dataset attribution
Wolberg, W., Mangasarian, O., Street, N., & Street, W. (1993). Breast Cancer Wisconsin (Diagnostic) [Dataset]. UCI Machine Learning Repository. DOI: 10.24432/C5DW2B. Dataset licensed CC BY 4.0. The benchmark data were not created by DeTection Inc.