Soham Bopardikar
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Soham Bopardikar has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Daniel Sierra-Sosa
Department of Computer Science, The Catholic University of America, Washington, DC, USA
2 shared publications - Alfaxad Eyembe
Faculty of Engineering, Kyoto University of Advanced Science (KUAS), Ukyo-ku, Kyoto 615-8577, Japan
1 shared publication Ashwini K
1 shared publicationAvinash D
1 shared publication- Begonya Garcia-Zapirain
University of Deusto
1 shared publication - Hemavathi Santhanam
Faculty of Graduate Studies and Research, Saint Mary’s University, 923 Robie Street, Halifax, NS B3H 3C3, Canada
1 shared publication
Latest Publications
- Open Access
- 6 February 2024
- Published by Springer Nature in Research Square
Quantum Machine Learning and TensorNetworks as an Aid in the Clinical Diagnosis ofCoronary Artery Disease
Current applications of quantum machine learning are emerging, dueto the potential benefits that quantum technologies could bring in thenear future. One of the most recent developments is tensor network-based architectures, to explore the feasibility of the application of thismethod to healthcare, in this paper tensor networks are applied to theIEEE Heart Disease (COMPREHENSIVE) dataset for supervised clas-sification of coronary artery disease diagnosis. Three quantum machinelearning models were implemented in this study: 3-qubit Q-MPS, 4 qubitQ-MERA and 4-qubit Q-TTN. These were compared to six classical ma-chine learning models: Logistic Regression, Naive Bayes, Support VectorMachines (SVM), Decision Tree, Random Forest, and XGBoost; achiev-ing similar or better results than those of the best classical models. Themodels were evaluated following different strategies to modify the trainingand testing conditions, applying variations to the dataset by isolating theCleveland and Hungary datasets, which are subsets of the former. Addi-tionally, the impact of the input feature space preparation is demonstratedexperimentally, showing that there exist preferred conditions where thegeneralization of the model is maximum.
- 6 February 2024
- Published by Springer Nature in Research Square
Open AccessQuantum Machine Learning and TensorNetworks as an Aid in the Clinical Diagnosis ofCoronary Artery Disease
Current applications of quantum machine learning are emerging, dueto the potential benefits that quantum technologies could bring in thenear future. One of the most recent developments is tensor network-based architectures, to explore the feasibility of the application of thismethod to healthcare, in this paper tensor networks are applied to theIEEE Heart Disease (COMPREHENSIVE) dataset for supervised clas-sification of coronary artery disease diagnosis. Three quantum machinelearning models were implemented in this study: 3-qubit Q-MPS, 4 qubitQ-MERA and 4-qubit Q-TTN. These were compared to six classical ma-chine learning models: Logistic Regression, Naive Bayes, Support VectorMachines (SVM), Decision Tree, Random Forest, and XGBoost; achiev-ing similar or better results than those of the best classical models. Themodels were evaluated following different strategies to modify the trainingand testing conditions, applying variations to the dataset by isolating theCleveland and Hungary datasets, which are subsets of the former. Addi-tionally, the impact of the input feature space preparation is demonstratedexperimentally, showing that there exist preferred conditions where thegeneralization of the model is maximum.
- Open Access
An approach to solve the coarse-grained Protein folding problem in a Quantum Computer
Protein folding, which dictates the protein structure from its amino acid sequence, is half a century old problem of biology. The function of the protein correlates with its structure, emphasizing the need of understanding protein folding for studying the cellular and molecular mechanisms that occur within biological systems. Understanding protein structures and enzymes plays a critical role in target based drug designing, elucidating protein-related disease mechanisms, and innovating novel enzymes. While recent advancements in AI based protein structure prediction methods have solved the protein folding problem to an extent, their precision in determining the structure of the protein with low sequence similarity is limited. Classical methods face challenges in generating extensive conformational samplings, making quantum-based approaches advantageous for solving protein folding problems. In this work we developed a novel turn based encoding algorithm that can be run on a gate based quantum computer for predicting the structure of smaller protein sequences using the HP model as an initial framework, which can be extrapolated in its application to larger and more intricate protein systems in future. The HP model best represents a major step in protein folding phenomena - the hydrophobic collapse which brings the hydrophobic amino acid to the interior of a protein. The folding problem is cast in a 3D cubic lattice with degrees of freedom along edges parallel to the orthogonal axes, as well as along diagonals parallel to the axial planes. While, the original formulation with higher order terms can be run on gate based quantum hardwares, the QUBO formulation can give results on both classical softwares employing annealers and IBM CPLEX as well as quantum hardwares.
Open AccessAn approach to solve the coarse-grained Protein folding problem in a Quantum Computer
Protein folding, which dictates the protein structure from its amino acid sequence, is half a century old problem of biology. The function of the protein correlates with its structure, emphasizing the need of understanding protein folding for studying the cellular and molecular mechanisms that occur within biological systems. Understanding protein structures and enzymes plays a critical role in target based drug designing, elucidating protein-related disease mechanisms, and innovating novel enzymes. While recent advancements in AI based protein structure prediction methods have solved the protein folding problem to an extent, their precision in determining the structure of the protein with low sequence similarity is limited. Classical methods face challenges in generating extensive conformational samplings, making quantum-based approaches advantageous for solving protein folding problems. In this work we developed a novel turn based encoding algorithm that can be run on a gate based quantum computer for predicting the structure of smaller protein sequences using the HP model as an initial framework, which can be extrapolated in its application to larger and more intricate protein systems in future. The HP model best represents a major step in protein folding phenomena - the hydrophobic collapse which brings the hydrophobic amino acid to the interior of a protein. The folding problem is cast in a 3D cubic lattice with degrees of freedom along edges parallel to the orthogonal axes, as well as along diagonals parallel to the axial planes. While, the original formulation with higher order terms can be run on gate based quantum hardwares, the QUBO formulation can give results on both classical softwares employing annealers and IBM CPLEX as well as quantum hardwares.
- Open Access
- 25 June 2023
- Published by MDPI AG in Journal of Imaging
- Vol. 9 (7) , 128
Hybrid Classical–Quantum Transfer Learning for Cardiomegaly Detection in Chest X-rays
Cardiovascular diseases are among the major health problems that are likely to benefit from promising developments in quantum machine learning for medical imaging. The chest X-ray (CXR), a widely used modality, can reveal cardiomegaly, even when performed primarily for a non-cardiological indication. Based on pre-trained DenseNet-121, we designed hybrid classical–quantum (CQ) transfer learning models to detect cardiomegaly in CXRs. Using Qiskit and PennyLane, we integrated a parameterized quantum circuit into a classic network implemented in PyTorch. We mined the CheXpert public repository to create a balanced dataset with 2436 posteroanterior CXRs from different patients distributed between cardiomegaly and the control. Using k-fold cross-validation, the CQ models were trained using a state vector simulator. The normalized global effective dimension allowed us to compare the trainability in the CQ models run on Qiskit. For prediction, ROC AUC scores up to 0.93 and accuracies up to 0.87 were achieved for several CQ models, rivaling the classical–classical (CC) model used as a reference. A trustworthy Grad-CAM++ heatmap with a hot zone covering the heart was visualized more often with the QC option than that with the CC option (94% vs. 61%, p < 0.001), which may boost the rate of acceptance by health professionals.
- 25 June 2023
- Published by MDPI AG in Journal of Imaging
- Vol. 9 (7) , 128
Open AccessHybrid Classical–Quantum Transfer Learning for Cardiomegaly Detection in Chest X-rays
Cardiovascular diseases are among the major health problems that are likely to benefit from promising developments in quantum machine learning for medical imaging. The chest X-ray (CXR), a widely used modality, can reveal cardiomegaly, even when performed primarily for a non-cardiological indication. Based on pre-trained DenseNet-121, we designed hybrid classical–quantum (CQ) transfer learning models to detect cardiomegaly in CXRs. Using Qiskit and PennyLane, we integrated a parameterized quantum circuit into a classic network implemented in PyTorch. We mined the CheXpert public repository to create a balanced dataset with 2436 posteroanterior CXRs from different patients distributed between cardiomegaly and the control. Using k-fold cross-validation, the CQ models were trained using a state vector simulator. The normalized global effective dimension allowed us to compare the trainability in the CQ models run on Qiskit. For prediction, ROC AUC scores up to 0.93 and accuracies up to 0.87 were achieved for several CQ models, rivaling the classical–classical (CC) model used as a reference. A trustworthy Grad-CAM++ heatmap with a hot zone covering the heart was visualized more often with the QC option than that with the CC option (94% vs. 61%, p < 0.001), which may boost the rate of acceptance by health professionals.