In the project Self-Organizing Convolutional Maps (SOCOM), drawing inspiration from both fields of deep learning and unsupervised learning, the study and development of an innovative self-organizing convolutional map (SOCOM) were proposed and implemented. SOCOM constitutes a hybridization of a purely unsupervised learning algorithm (the Self-Organizing Map, SOM) with the core structure of a deep learning algorithm (the Convolutional Neural Network, CNN). A novel and innovative hybrid architecture was developed integrating, into a unified and indivisible architectural and algorithmic framework, the discovery/recognition of representations and patterns in the input data 
space with clustering functionalities in the output space. The ultimate goal is to create a model capable of providing clustering and unsupervised categorization capabilities by utilizing high-level abstract representations. The proposed model also featured visualization capabilities that are leveraged as tools for extracting information, generating new knowledge, understanding the internal operations, and interpreting the relationships between the results. To the best of our knowledge, the proposed architecture had not been proposed in the literature. Furthermore, the effectiveness of the proposed model was validated by applying it to a wide range of publicly available time-series datasets. During the project, a presentation at an international scientific conference and a publication in a peer-reviewed international scientific journal were delivered.
Project Τitle: Self-Organizing Convolutional Maps (SOCOM)
Project Duration: 01/04/2020 – 31/01/2022
Project Framework & Funding: Partnership Agreement for the Development Framework (PA) 2014-2020 – 50.050,00 EUR
Scientific coordinator (UNIWA): Stelios MITILINEOS – Professor, EEE Department, UNIWa
Research Publications / Results / Patents:
[1] Ferles, C., Papanikolaou, Y., Savvaidis, S. P., and Mitilineos, S. A., “Deep self-organizing map of convolutional layers for clustering and visualizing image data”, Machine Learning and Knowledge Extraction, Vol. 3, No. 4, pp. 879-899, November 2021; DOI: 10.3390/make3040044.
[2] Ferles, C., Papanikolaou, Y., Savvaidis, S. P., and Mitilineos, S. A., “Deep learning self-organizing map of convolutional layers”, Proceedings of the 2nd International Conference on Artificial Intelligence and Big Data (AIBD 2021), pp. 1-8, March 20-21, 2021, Vienna, AUSTRIA.
