Research
(Also see Publications)
Our research focuses on developing quantum computing methods applicable to solving problems in science, engineering and business. We investigate quantum machine learning models capable of representing and processing complex information structures. We also study quantum optimization techniques working effectively in highly constrained problem domains. All our work is verified experimentally to ensure smooth transition from the lab to practice.
Recent projects
Quantum model expressivity and trainability by Jacob L. Cybulski and Sebastian Zając. The project investigates the quantum model attributes that promote its ability to effectively represent
classical data in quantum space, and often conflicting attributes that promote its capacity to learn and generalize in the process of optimization.
Architectural design of quantum time series Autoencoders by Jacob L. Cybulski, Sebastian Zając, Jacob Zwoniarski and Artur Strąg. This project explores various approaches to designing quantum autoencoders, to improve their training efficiency and their ability to remove noise from the represented time series data.
Exploration of quantum reservoir computing (QRC) models by Jacob L. Cybulski, Staszek Władyka and Sebastian Zając. The focus of this project is on the exploration of novel approaches to enhancing classical reservoir computing with quantum techniques. One of the surprising outcomes of such a fusion is the emergence of dual-memory reservoir systems, able to accomodate short- and long-term memory of the resulting predictive system.
Quantum encoding and decoding of time series and their impact on the model effectiveness by Jacob L. Cybulski and Sebastian Zając. This project investigates two aspects of quantum model development that are often not given sufficient priority, i.e. quantum encoding of classical information, as well as measurement of the model state and the subsequent interpretation of measurement results. The research also provides advice for quantum researchers and practitioners.
Quantum graphs by Sebastian Zając and Jacob L. Cybulski. There are very few attempts at quantum representation and processing of graphs. In this project we are looking to utilize Hilbert space to gain efficiency of processing dynamic complex graphs, such as those representing social media interactions and information flow.
Quantum time series analysis by Jacob L. Cybulski and Sebastian Zając. This is one of the ongoing projects, with many facets and directions, including using quantum machine learning techniques for time series prediction, noise reduction in signals and complex anomaly detection in univariate and multivariate temporal data.
Study of barren plateaus and quantum models’ effective dimension by Jacob L. Cybulski and Thanh (Tim) Nguyen. Training of quantum neural networks (QNN) often faces difficulties due to the emergence of barren plateaus (BP), i.e. large flat areas in the gradient of loss function, which prevent optimization of model parameters. This project evaluated four approaches to BP mitigation and proposed a way of assessing their effectiveness by relying on the measurement of the model effective dimension.
