The project focuses on the application of the Parrondo paradox - a phenomenon in which the combination of two losing strategies leads to a favorable outcome - in machine learning processes. The goal of the research is to develop new optimization mechanisms inspired by effects observed in quantum models, such as quantum random walks. Particular emphasis will be placed on studying information propagation in complex systems and adapting these mechanisms to classical learning problems. A unique aspect of the project is the use of Parrondo's paradox in the field of machine learning, which makes the applicant's research different from other projects he has undertaken. Quantum random error models are a key inspiration for the project. In both discrete and continuous variants, the research has shown that the alternating application of actions of different nature - each of which individually leads to an unfavorable outcome - can paradoxically contribute to more efficient information propagation. Using inspiration from models of quantum random erring can lead to the development of new algorithms that process information more efficiently, especially in tasks that require modeling complex relationships between data elements, especially applications of these techniques in the context of visual data analysis and processing. The development of a new theoretical framework and its experimental implementation will open up the possibility of developing advanced learning algorithms that can be applied not only in computer vision, but also in other areas requiring the processing of high-dimensional data.