Publications
Banner CSBC2026: dc-qml: data-centric quantum machine learning
Year: 2026
Authors: E. S. Flores , K. D. R. C. Barbosa, B. V. Guterres S. S. C. Botelho, M. R. Pias
Summary:This poster presents the dc-qml (data-centric quantum machine learning) methodology, which integrates data preparation, encoding, and representation into a unified workflow for quantum machine learning. The work highlights the importance of data-centric approaches to improve the performance of near-term quantum systems and presents preliminary results for industrial anomaly detection. The poster was presented at the **1st Brazilian Symposium on Quantum Computing and Communication (SBCCQ 2026), held on July 23, 2026, in Gramado, Rio Grande do Sul, Brazil.
PDF |
Data-Centric Quantum Machine Learning: data-to-models continuum
Year: 2026
Authors: Dalla Riva Cucco Barbosa, Kauã; Flores, Everson; Guterres, Bruna; Silva da Costa Botelho, Silvia; Rita Pias, Marcelo
Summary:This work presents a review of data-centric approaches in Quantum Machine Learning (QML). The study examines data preparation and representation requirements across different QML architectures and proposes a taxonomy based on dimensionality reduction, data encoding, balancing, and compression. Building on this analysis, the **dc-qml** framework is introduced, integrating stages such as data ingestion, abstraction, tokenization, quantum encoding, and synthetic data generation. The results indicate that improvements in the performance of quantum systems are often associated with data preparation strategies, highlighting the importance of data-centric approaches for for the field of Quantum Machine Learning.
dc-qml: data-centric quantum machine learning
Year: 2026
Authors: Dalla Riva Cucco Barbosa, Kauã; Flores, Everson; Guterres, Bruna; Silva da Costa Botelho, Silvia; Rita Pias, Marcelo
Summary:This work presents the **dc-qml** methodology, a data-centric approach to Quantum Machine Learning. The proposed framework organizes the stages of data ingestion, preparation, encoding, and synthetic data generation into an integrated workflow for training quantum models. Unlike traditional approaches, which primarily focus on the development of algorithms and quantum circuits, the methodology emphasizes data quality and representation as key factors influencing the performance of QML systems. Preliminary results highlight the potential of this approach to support the development of more efficient quantum applications.
Unsupervised Quantum Kernels for One-Class Fault Detection in Industrial Time Series
Year: 2026
Authors: Dalla Riva Cucco Barbosa, Kauã; Flores, Everson; Guterres, Bruna; Silva da Costa Botelho, Silvia; Rita Pias, Marcelo
Summary:This work evaluates the use of quantum kernels for anomaly detection in industrial time series with limited availability of labeled data. The OC-SVM and OC-QSVM models were compared across ten industrial scenarios using the same preprocessing and training procedures. The results show that both models achieve similar performance in simpler cases, while more significant differences emerge in scenarios involving complex multivariate patterns. The OC-QSVM model achieved the best performance in one of the evaluated scenarios, suggesting that quantum feature spaces can capture relationships between variables that are not easily identified by classical approaches.
Data-centric Sensitivity Analysis of Quantum Kernels for Industrial Time Series
Year: 2026
Authors: Dalla Riva Cucco Barbosa, Kauã; Flores, Everson; Guterres, Bruna; Silva da Costa Botelho, Silvia; Rita Pias, Marcelo
Summary:This work investigates the behavior of classical and quantum kernels for anomaly detection in industrial environments. SVM models using the RBF kernel were compared with QSVM models using the qIT quantum kernel under different encoding resolutions and training data sizes. The results show that model performance depends on data representation conditions, with each kernel exhibiting distinct behavior. The qIT quantum kernel demonstrated greater stability across the evaluated configurations, indicating its potential for industrial Quantum Machine Learning applications.
Data-Centric Evaluation of Quantum Machine Learning for Industrial Fault Detection
Year: 2026
Authors: Dalla Riva Cucco Barbosa, Kauã; Flores, Everson; Guterres, Bruna; Silva da Costa Botelho, Silvia; Rita Pias, Marcelo
Summary:This work investigates the use of Quantum Machine Learning for industrial fault detection from a data-centric perspective. QSVM and SVM models were evaluated under different data partitioning protocols while keeping the preprocessing and encoding procedures unchanged. The results show that QSVM performance is more sensitive to the way data are partitioned, whereas SVM exhibits more stable behavior. The study highlights the importance of data partitioning in the evaluation of Quantum Machine Learning solutions.