Dr hab. Karol Bartkiewicz, prof. UAM
Department of Nonlinear Optics
- Tel: +48 61 829 5236
- Loc: wing G, second floor, room 279
- Email: bark@amu.edu.pl
- URL: http://bark.home.amu.edu.pl/
Scientific degrees
Habilitation – 2019
PhD in physics – 2012
MSc in physics – 2008
Research interests
The scope of my scientific interests includes quantum physics and optics, more specifically:
- quantum cloning,
- quantum teleportation,
- quantum cryptography,
- quantum correlations,
- machine learning,
- optical methods of quantum information processing,
- mathematical and computational methods of physics,
- foudations of physics,
- photonics and quantum optics.
Projects
| 1. | Karol Bartkiewicz Kernel based quantum machine learning in optical circuits 2019 - 2021, (Czech Science Foundation (at Palacy University in Olomouc), budget: 4 875 000 CZK (~ 850 000 PLN)). @misc{Bartkiewicz2021, title = {Kernel based quantum machine learning in optical circuits}, author = {Karol Bartkiewicz}, year = {2021}, date = {2021-12-01}, abstract = {This project focuses on theoretical and experimental research on kernel-based quantum machine learning (QML) using linear optics and individual photons as information carriers. QML is crucial for future development of quantum artificial neural networks and other quantum-enhanced technologies. QML can exponentially improve machine learning (ML) vastly applied in many industries. Here, we focus on three research objectives.First task is to design quantum optical circuits and implementing assorted kernels for QML. We plan on implementing these kernels to verify the benefits of this approach to QML over other solutions described in the literature.Second part of the project is to apply QML to classification or clustering, i.e., typical supervised and unsupervised ML problems. We plan to apply QML to learning properties of a quantum system. We will be developing a fremework for explaining parameters of QML system in terms directly relevant to the investigated problem. In the third stage we plan on applying QML to generative models (creating new data based on the training data). Three research objectives will be investigated both theoretically and experimentally: (i) quantum optical circuits for QML kernel implementation, (ii) applications and explainability of QML with kernels for supervised and unsupervised learning problems, (iii) quantum generative models.}, howpublished = {2019}, note = {Czech Science Foundation (at Palacy University in Olomouc), budget: 4 875 000 CZK (~ 850 000 PLN)}, keywords = {}, pubstate = {published}, tppubtype = {misc} } This project focuses on theoretical and experimental research on kernel-based quantum machine learning (QML) using linear optics and individual photons as information carriers. QML is crucial for future development of quantum artificial neural networks and other quantum-enhanced technologies. QML can exponentially improve machine learning (ML) vastly applied in many industries. Here, we focus on three research objectives.First task is to design quantum optical circuits and implementing assorted kernels for QML. We plan on implementing these kernels to verify the benefits of this approach to QML over other solutions described in the literature.Second part of the project is to apply QML to classification or clustering, i.e., typical supervised and unsupervised ML problems. We plan to apply QML to learning properties of a quantum system. We will be developing a fremework for explaining parameters of QML system in terms directly relevant to the investigated problem. In the third stage we plan on applying QML to generative models (creating new data based on the training data). Three research objectives will be investigated both theoretically and experimentally: (i) quantum optical circuits for QML kernel implementation, (ii) applications and explainability of QML with kernels for supervised and unsupervised learning problems, (iii) quantum generative models. |
Publications
2022 |
|
| 3. | Jan Roik, Karol Bartkiewicz, Antonín Černoch, Karel Lemr Entanglement quantification from collective measurements processed by machine learning Physics Letters A, 446 , pp. 128270, 2022, ISSN: 0375-9601. @article{ROIK2022128270b, title = {Entanglement quantification from collective measurements processed by machine learning}, author = {Jan Roik and Karol Bartkiewicz and Antonín Černoch and Karel Lemr}, url = {https://www.sciencedirect.com/science/article/pii/S0375960122003528}, doi = {https://doi.org/10.1016/j.physleta.2022.128270}, issn = {0375-9601}, year = {2022}, date = {2022-09-15}, journal = {Physics Letters A}, volume = {446}, pages = {128270}, abstract = {This paper investigates how to reduce the number of measurement configurations needed for sufficiently precise entanglement quantification. Instead of analytical formulae, we employ artificial neural networks to predict the amount of entanglement in a quantum state based on results of collective measurements (simultaneous measurements on multiple instances of the investigated state). We consider collective measurement limited to two copies of the investigated state. This approach allows us to explore the precision of entanglement quantification as a function of measurement configurations in a relevant scenario for practical quantum communications. For the purpose of our research, we consider general two-qubit states and their negativity as entanglement quantifier. We outline the benefits of this approach in future quantum communication networks.}, keywords = {}, pubstate = {published}, tppubtype = {article} } This paper investigates how to reduce the number of measurement configurations needed for sufficiently precise entanglement quantification. Instead of analytical formulae, we employ artificial neural networks to predict the amount of entanglement in a quantum state based on results of collective measurements (simultaneous measurements on multiple instances of the investigated state). We consider collective measurement limited to two copies of the investigated state. This approach allows us to explore the precision of entanglement quantification as a function of measurement configurations in a relevant scenario for practical quantum communications. For the purpose of our research, we consider general two-qubit states and their negativity as entanglement quantifier. We outline the benefits of this approach in future quantum communication networks. |
2021 |
|
| 2. | Kateřina Jirákov á, Antonín Č, Karel Lemr, Karol Bartkiewicz, Adam Miranowicz Physical Review A, 104 (6), pp. 062436, 2021. @article{Jirakova2021b, title = {Experimental hierarchy and optimal robustness of quantum correlations of two-qubit states with controllable white noise}, author = {Kate{ř}ina Jirákov á and Antonín Č and Karel Lemr and Karol Bartkiewicz and Adam Miranowicz}, url = {https://doi.org/10.1103/physreva.104.062436}, doi = {10.1103/physreva.104.062436}, year = {2021}, date = {2021-12-21}, journal = {Physical Review A}, volume = {104}, number = {6}, pages = {062436}, publisher = {American Physical Society (APS)}, abstract = {We demonstrate a hierarchy of various classes of quantum correlations on experimentally prepared two-qubit Werner-like states with controllable white noise. Werner states, which are white-noise-affected Bell states, are prototypal examples for studying such a hierarchy as a function of the amount of white noise. We experimentally generate Werner states and their generalizations, i.e., partially entangled pure states affected by white noise. These states enable us to study the hierarchy of the following classes of correlations: separability, entanglement, steering in three- and two-measurement scenarios, and Bell nonlocality. We show that the generalized Werner states (GWSs) reveal fundamentally different aspects of the hierarchy compared to the Werner states. In particular, we find five different parameter regimes of the GWSs, including those steerable in a two-measurement scenario but not violating Bell inequalities. This regime cannot be observed for the usual Werner states. Moreover, we find threshold curves separating different regimes of the quantum correlations and find the optimal states which allow for the largest amount of white noise which does not destroy their specific quantum correlations (e.g., unsteerable entanglement). Thus, we could identify the optimal Bell-nondiagonal GWSs which are, for this specific meaning, more robust against the white noise compared to the Bell-diagonal GWSs (i.e., Werner states).}, keywords = {}, pubstate = {published}, tppubtype = {article} } We demonstrate a hierarchy of various classes of quantum correlations on experimentally prepared two-qubit Werner-like states with controllable white noise. Werner states, which are white-noise-affected Bell states, are prototypal examples for studying such a hierarchy as a function of the amount of white noise. We experimentally generate Werner states and their generalizations, i.e., partially entangled pure states affected by white noise. These states enable us to study the hierarchy of the following classes of correlations: separability, entanglement, steering in three- and two-measurement scenarios, and Bell nonlocality. We show that the generalized Werner states (GWSs) reveal fundamentally different aspects of the hierarchy compared to the Werner states. In particular, we find five different parameter regimes of the GWSs, including those steerable in a two-measurement scenario but not violating Bell inequalities. This regime cannot be observed for the usual Werner states. Moreover, we find threshold curves separating different regimes of the quantum correlations and find the optimal states which allow for the largest amount of white noise which does not destroy their specific quantum correlations (e.g., unsteerable entanglement). Thus, we could identify the optimal Bell-nondiagonal GWSs which are, for this specific meaning, more robust against the white noise compared to the Bell-diagonal GWSs (i.e., Werner states). |
| 1. | Jan Roik, Karol Bartkiewicz, Antonín Černoch, Karel Lemr Phys. Rev. Applied, 15 , pp. 054006, 2021. @article{Bartkiewicz2021b, title = {Accuracy of Entanglement Detection via Artificial Neural Networks and Human-Designed Entanglement Witnesses}, author = {Jan Roik and Karol Bartkiewicz and Antonín Černoch and Karel Lemr}, url = {https://journals.aps.org/prapplied/abstract/10.1103/PhysRevApplied.15.054006}, doi = {https://doi.org/10.1103/PhysRevApplied.15.054006}, year = {2021}, date = {2021-05-04}, journal = {Phys. Rev. Applied}, volume = {15}, pages = {054006}, abstract = {The detection of entangled states is essential in both fundamental and applied quantum physics. However, this task proves to be challenging, especially for general quantum states. One can execute full state tomography but this method is time demanding, especially in complex systems. Other approaches use entanglement witnesses: these methods tend to be less demanding but lack reliability. Here, we demonstrate that artificial neural networks (ANNs) provide a balance between the two approaches. In this paper, we make a comparison of ANN performance with witness-based methods for random general two-qubit quantum states without any prior information on the states. Furthermore, we apply our approach to a real experimental data set.}, keywords = {}, pubstate = {published}, tppubtype = {article} } The detection of entangled states is essential in both fundamental and applied quantum physics. However, this task proves to be challenging, especially for general quantum states. One can execute full state tomography but this method is time demanding, especially in complex systems. Other approaches use entanglement witnesses: these methods tend to be less demanding but lack reliability. Here, we demonstrate that artificial neural networks (ANNs) provide a balance between the two approaches. In this paper, we make a comparison of ANN performance with witness-based methods for random general two-qubit quantum states without any prior information on the states. Furthermore, we apply our approach to a real experimental data set. |