Aikaterini Mandilara
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
Aikaterini Mandilara has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Dimitris Syvridis
National and Kapodistrian University of Athens
7 shared publications - Georgios Maragkopoulos
Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Greece
4 shared publications - Antonia Tsili
Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Panepistimioupolis, Ilisia, 16122 Athens, Greece
3 shared publications - Vladimir M. Akulin
Laboratoire Aimé Cotton, CNRS (UPR 3321), Bâtiment 505, 91405 Orsay Cedex, France
2 shared publications A. Gabdulin
Department of Physics, School of Science and Humanities, Nazarbayev University, 53 Qabanbay Batyr Avenue, Nur-Sultan 010000, Kazakhstan
1 shared publicationAristeides D Papadopoulos
Eulambia Advanced technologies Ltd
1 shared publication
Latest Publications
- Open Access
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 035061
Learning quantum kernels with continuous-variable optical circuits
Kernel-based learning methods are central to classical machine learning, yet their quantum counterparts have largely remained limited to fixed circuit constructions. Most existing approaches aim to improve the performance of kernel-based support vector machines by heuristically tailoring kernel functions to specific tasks. In this work, we introduce a principled framework for adapting quantum kernels by translating two well-established classical techniques—the Fisher criterion and quasi-conformal kernel transformations—into the setting of continuous-variable (CV) quantum circuits. We show that the classical Fisher criterion is equivalent to maximizing the Hilbert–Schmidt distance in the quantum feature space induced by a quantum circuit. We further demonstrate that quasi-conformal kernel transformations admit a direct and natural implementation at the circuit level, realized by augmenting the original quantum circuit with additional modes. Using experimentally accessible optical circuits composed solely of displacement and squeezing operations, we numerically demonstrate that each method independently leads to a systematic enhancement of learning performance. Taken together, our theoretical and numerical results establish a concrete route for importing powerful classical kernel-learning strategies into quantum machine learning, and position CV quantum optics as a flexible and experimentally viable platform for adaptive, task-driven quantum kernel design.
- 1 June 2026
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 7 (3) , 035061
Open AccessLearning quantum kernels with continuous-variable optical circuits
Kernel-based learning methods are central to classical machine learning, yet their quantum counterparts have largely remained limited to fixed circuit constructions. Most existing approaches aim to improve the performance of kernel-based support vector machines by heuristically tailoring kernel functions to specific tasks. In this work, we introduce a principled framework for adapting quantum kernels by translating two well-established classical techniques—the Fisher criterion and quasi-conformal kernel transformations—into the setting of continuous-variable (CV) quantum circuits. We show that the classical Fisher criterion is equivalent to maximizing the Hilbert–Schmidt distance in the quantum feature space induced by a quantum circuit. We further demonstrate that quasi-conformal kernel transformations admit a direct and natural implementation at the circuit level, realized by augmenting the original quantum circuit with additional modes. Using experimentally accessible optical circuits composed solely of displacement and squeezing operations, we numerically demonstrate that each method independently leads to a systematic enhancement of learning performance. Taken together, our theoretical and numerical results establish a concrete route for importing powerful classical kernel-learning strategies into quantum machine learning, and position CV quantum optics as a flexible and experimentally viable platform for adaptive, task-driven quantum kernel design.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19, 29275-29292
Parameterized Quantum Circuits as Feature Maps: Representation Quality and Readout Effects in Multispectral Land-Cover Classification
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 19, 29275-29292
Open AccessParameterized Quantum Circuits as Feature Maps: Representation Quality and Readout Effects in Multispectral Land-Cover Classification
- Open Access
- 30 July 2025
- Published by Springer Nature in Quantum Information Processing
- Vol. 24 (8) , 1-17
Enhancing the performance of variational quantum classifiers with hybrid autoencoders
Variational quantum circuits (VQC) lie at the forefront of quantum machine learning research. Still, the use of quantum networks for real data processing remains challenging as the number of available qubits cannot accommodate a large dimensionality of data—if the usual angle encoding scenario is used. To achieve dimensionality reduction, Principal Component Analysis is routinely applied as a pre-processing method before the embedding of the classical features on qubits. In this work, we propose an alternative method which reduces the dimensionality of a given dataset by taking into account the specific quantum embedding that comes after. This method aspires to make quantum machine learning with VQCs more versatile and effective on datasets of high dimension. At a second step, we propose a quantum-inspired classical autoencoder model which can be used to encode information in low latent spaces. The power of our proposed models is exhibited via numerical tests. We show that our targeted dimensionality reduction method considerably boosts VQC’s performance, and we also identify cases for which the second model outperforms classical autoencoders in terms of reconstruction loss.
- 30 July 2025
- Published by Springer Nature in Quantum Information Processing
- Vol. 24 (8) , 1-17
Open AccessEnhancing the performance of variational quantum classifiers with hybrid autoencoders
Variational quantum circuits (VQC) lie at the forefront of quantum machine learning research. Still, the use of quantum networks for real data processing remains challenging as the number of available qubits cannot accommodate a large dimensionality of data—if the usual angle encoding scenario is used. To achieve dimensionality reduction, Principal Component Analysis is routinely applied as a pre-processing method before the embedding of the classical features on qubits. In this work, we propose an alternative method which reduces the dimensionality of a given dataset by taking into account the specific quantum embedding that comes after. This method aspires to make quantum machine learning with VQCs more versatile and effective on datasets of high dimension. At a second step, we propose a quantum-inspired classical autoencoder model which can be used to encode information in low latent spaces. The power of our proposed models is exhibited via numerical tests. We show that our targeted dimensionality reduction method considerably boosts VQC’s performance, and we also identify cases for which the second model outperforms classical autoencoders in terms of reconstruction loss.
- Open Access
- 25 February 2025
- Published by MDPI AG in Electronics
- Vol. 14 (5) , 901
Learning the N-Input Parity Function with a Single-Qubit and Single-Measurement Sampling
The parity problem, a generalization of the XOR problem to higher-dimensional inputs, is a challenging benchmark for evaluating learning algorithms, due to its increased complexity as the number of dimensions of the feature space grows. In this work, a single-qubit classifier is developed, which can efficiently learn the parity function from input data. Despite the qubit model’s simplicity, the solution landscape created in the context of the parity problem offers an attractive test bed for exploring optimization methods for quantum classifiers. We propose a new optimization method called Ensemble Stochastic Gradient Descent (ESGD), with which density matrices describing batches of quantum states are incorporated into the loss function. We demonstrate that ESGD outperforms both Gradient Descent and Stochastic Gradient Descent given the aforementioned problem. Additionally, we show that applying ESGD with only one measurement per data input does not lead to any performance degradation. Our findings not only highlight the potential of a single-qubit model, but also offer valuable insights into the use of density matrices for optimization. Further to this, we complement the outcome with interesting results arising by the employment of a Doubly Stochastic Gradient Descent for training quantum variational circuits.
- 25 February 2025
- Published by MDPI AG in Electronics
- Vol. 14 (5) , 901
Open AccessLearning the N-Input Parity Function with a Single-Qubit and Single-Measurement Sampling
The parity problem, a generalization of the XOR problem to higher-dimensional inputs, is a challenging benchmark for evaluating learning algorithms, due to its increased complexity as the number of dimensions of the feature space grows. In this work, a single-qubit classifier is developed, which can efficiently learn the parity function from input data. Despite the qubit model’s simplicity, the solution landscape created in the context of the parity problem offers an attractive test bed for exploring optimization methods for quantum classifiers. We propose a new optimization method called Ensemble Stochastic Gradient Descent (ESGD), with which density matrices describing batches of quantum states are incorporated into the loss function. We demonstrate that ESGD outperforms both Gradient Descent and Stochastic Gradient Descent given the aforementioned problem. Additionally, we show that applying ESGD with only one measurement per data input does not lead to any performance degradation. Our findings not only highlight the potential of a single-qubit model, but also offer valuable insights into the use of density matrices for optimization. Further to this, we complement the outcome with interesting results arising by the employment of a Doubly Stochastic Gradient Descent for training quantum variational circuits.
- Open Access
- 29 January 2025
- Published by MDPI AG in Preprints.org
Learning the N-input Parity Function with a Single Qubit and Single Measurement Sampling
The parity problem, a generalization of the XOR problem to higher-dimensional inputs, is a challenging benchmark for evaluating learning algorithms, due to its increased complexity as the number of dimensions of the feature space grows. In this work, a single-qubit classifier is developed, which can efficiently learn the parity function from input data. Despite the qubit model’s simplicity, the solution landscape created in the context of the parity problem offers an attractive test bed for exploring optimization methods for quantum classifiers. We propose a new optimization method called Ensemble Stochastic Gradient Descent (ESGD), with which density matrices describing batches of quantum states are incorporated into the loss function. We demonstrate that ESGD outperforms both Gradient Descent and Stochastic Gradient Descent given the aforementioned problem. Additionally, we show that applying ESGD with only one measurement per data input does not lead to any performance degradation. Our findings not only highlight the potential of a single qubit model, but also offer valuable insights into the use of density matrices for optimization. Further to this, we complement the outcome with interesting results arising by the employment of a doubly stochastic gradient descent for training quantum variational circuits.
- 29 January 2025
- Published by MDPI AG in Preprints.org
Open AccessLearning the N-input Parity Function with a Single Qubit and Single Measurement Sampling
The parity problem, a generalization of the XOR problem to higher-dimensional inputs, is a challenging benchmark for evaluating learning algorithms, due to its increased complexity as the number of dimensions of the feature space grows. In this work, a single-qubit classifier is developed, which can efficiently learn the parity function from input data. Despite the qubit model’s simplicity, the solution landscape created in the context of the parity problem offers an attractive test bed for exploring optimization methods for quantum classifiers. We propose a new optimization method called Ensemble Stochastic Gradient Descent (ESGD), with which density matrices describing batches of quantum states are incorporated into the loss function. We demonstrate that ESGD outperforms both Gradient Descent and Stochastic Gradient Descent given the aforementioned problem. Additionally, we show that applying ESGD with only one measurement per data input does not lead to any performance degradation. Our findings not only highlight the potential of a single qubit model, but also offer valuable insights into the use of density matrices for optimization. Further to this, we complement the outcome with interesting results arising by the employment of a doubly stochastic gradient descent for training quantum variational circuits.
- Open Access
Real-Time Diagnostics on a QKD Link via QBER Time-Series Analysis
The integration of QKD systems in metro optical networks raises challenges that cannot be fully resolved with current technological means. In this work, we devised a methodology for identifying different types of impairments for a QKD link embedded in a communication network. Identification occurs in real time using a supervised machine learning model designed for this purpose. The model takes only QBER and SKR time-series data as the input, making its applicability not restricted to any specific QKD protocol or system. The output of the model specifies the working conditions for the QKD link, which is information that can be valuable for users and key management systems.
Open AccessReal-Time Diagnostics on a QKD Link via QBER Time-Series Analysis
The integration of QKD systems in metro optical networks raises challenges that cannot be fully resolved with current technological means. In this work, we devised a methodology for identifying different types of impairments for a QKD link embedded in a communication network. Identification occurs in real time using a supervised machine learning model designed for this purpose. The model takes only QBER and SKR time-series data as the input, making its applicability not restricted to any specific QKD protocol or system. The output of the model specifies the working conditions for the QKD link, which is information that can be valuable for users and key management systems.
- Open Access
- 7 March 2024
- Published by Springer Nature in Quantum Machine Intelligence
- Vol. 6 (1) , 1-13
Classification of data with a qudit, a geometric approach
We propose a model for data classification using isolated quantum $$\varvec{d}$$ -level systems or else qudits. The procedure consists of an encoding phase where classical data are mapped on the surface of the qudit’s Bloch hyper-sphere via rotation encoding, followed by a rotation of the sphere and a projective measurement. The rotation is adjustable in order to control the operator to be measured, while additional weights are introduced in the encoding phase adjusting the mapping on the Bloch’s hyper-surface. During the training phase, a cost function based on the average expectation value of the observable is minimized using gradient descent thereby adjusting the weights. Using examples and performing a numerical estimation of lossless memory dimension, we demonstrate that this geometrically inspired qudit model for classification is able to solve nonlinear classification problems using a small number of parameters only and without requiring entangling operations.- 7 March 2024
- Published by Springer Nature in Quantum Machine Intelligence
- Vol. 6 (1) , 1-13
Open AccessClassification of data with a qudit, a geometric approach
We propose a model for data classification using isolated quantum $$\varvec{d}$$ -level systems or else qudits. The procedure consists of an encoding phase where classical data are mapped on the surface of the qudit’s Bloch hyper-sphere via rotation encoding, followed by a rotation of the sphere and a projective measurement. The rotation is adjustable in order to control the operator to be measured, while additional weights are introduced in the encoding phase adjusting the mapping on the Bloch’s hyper-surface. During the training phase, a cost function based on the average expectation value of the observable is minimized using gradient descent thereby adjusting the weights. Using examples and performing a numerical estimation of lossless memory dimension, we demonstrate that this geometrically inspired qudit model for classification is able to solve nonlinear classification problems using a small number of parameters only and without requiring entangling operations. - Open Access
The Gell-Mann feature map of qutrits and its applications in classification tasks
Recent advancements in quantum hardware have enabled the realization of high-dimensional quantum states. This work investigates the potential of qutrits in quantum machine learning, leveraging their larger state space for enhanced supervised learning tasks. To that end, the Gell-Mann feature map is introduced which encodes information within an $8$-dimensional Hilbert space. The study focuses on classification problems, comparing Gell-Mann feature map with maps generated by established qubit and classical models. We test different circuit architectures and explore possibilities in optimization techniques. By shedding light on the capabilities and limitations of qutrit-based systems, this research aims to advance applications of low-depth quantum circuits.
Open AccessThe Gell-Mann feature map of qutrits and its applications in classification tasks
Recent advancements in quantum hardware have enabled the realization of high-dimensional quantum states. This work investigates the potential of qutrits in quantum machine learning, leveraging their larger state space for enhanced supervised learning tasks. To that end, the Gell-Mann feature map is introduced which encodes information within an $8$-dimensional Hilbert space. The study focuses on classification problems, comparing Gell-Mann feature map with maps generated by established qubit and classical models. We test different circuit architectures and explore possibilities in optimization techniques. By shedding light on the capabilities and limitations of qutrit-based systems, this research aims to advance applications of low-depth quantum circuits.
- 13 May 2021
- Published by IOP Publishing in Journal of Physics B: Atomic, Molecular and Optical Physics
- Vol. 54 (10) , 105501
Detecting the event of a single photon loss on quantum signals
We design a scheme for detecting a single photon loss from multi-modal quantum signals transmitted via a fiber or in free space. This consists of a special type of unitary coding transformation, the parity controlled-squeezing, applied prior to the transmission on the signal composed by information and ancilla modes. At the receiver, the inverse unitary transformation is applied—decoding, and the ancilla mode is measured via photon detection. The outcome reveals whether a photon loss has occurred. Distortion of the information part of the signal caused by an ancilla photon loss can be corrected via unitary transformation while loss of a photon from the information part of the signal can be detected with the probability exponentially close to unity but cannot be corrected. In contrast to the schemes of decoherence free subspaces and quantum error correction protocols, this method allows one to make use in principle of entire Hilbert space dimensionality. We discuss possible ways of synthesizing the required encoding–decoding transformation.
- 13 May 2021
- Published by IOP Publishing in Journal of Physics B: Atomic, Molecular and Optical Physics
- Vol. 54 (10) , 105501
Detecting the event of a single photon loss on quantum signals
We design a scheme for detecting a single photon loss from multi-modal quantum signals transmitted via a fiber or in free space. This consists of a special type of unitary coding transformation, the parity controlled-squeezing, applied prior to the transmission on the signal composed by information and ancilla modes. At the receiver, the inverse unitary transformation is applied—decoding, and the ancilla mode is measured via photon detection. The outcome reveals whether a photon loss has occurred. Distortion of the information part of the signal caused by an ancilla photon loss can be corrected via unitary transformation while loss of a photon from the information part of the signal can be detected with the probability exponentially close to unity but cannot be corrected. In contrast to the schemes of decoherence free subspaces and quantum error correction protocols, this method allows one to make use in principle of entire Hilbert space dimensionality. We discuss possible ways of synthesizing the required encoding–decoding transformation.
- 18 December 2019
- Published by American Physical Society (APS) in Physical Review A
- Vol. 100 (6) , 062322
Investigating bound entangled two-qutrit states via the best separable approximation
We use the linear programming algorithm introduced by Akulin et al. [V. M. Akulin, G. A. Kabatiansky, and A. Mandilara, Phys. Rev. A 92, 042322 (2015)] to perform best separable approximation on two-qutrit random density matrices. We combine the numerical results with theoretical methods in order to generate random representative families of positive partial transposed bound entangled (BE) states and analyze their properties. Our results disclose that for the two-qutrit system the BE states have negligible volume and that these form tiny “islands” sporadically distributed over the surface of the polytope of separable states. The detected families of BE states are found to be located under a layer of pseudo-one-copy undistillable negative partial transposed states with the latter covering the vast majority of the surface of the separable polytope.
- 18 December 2019
- Published by American Physical Society (APS) in Physical Review A
- Vol. 100 (6) , 062322
Investigating bound entangled two-qutrit states via the best separable approximation
We use the linear programming algorithm introduced by Akulin et al. [V. M. Akulin, G. A. Kabatiansky, and A. Mandilara, Phys. Rev. A 92, 042322 (2015)] to perform best separable approximation on two-qutrit random density matrices. We combine the numerical results with theoretical methods in order to generate random representative families of positive partial transposed bound entangled (BE) states and analyze their properties. Our results disclose that for the two-qutrit system the BE states have negligible volume and that these form tiny “islands” sporadically distributed over the surface of the polytope of separable states. The detected families of BE states are found to be located under a layer of pseudo-one-copy undistillable negative partial transposed states with the latter covering the vast majority of the surface of the separable polytope.