Dimitris Petratos

I received my Bachelors' diploma from the Department of Mathematics, University of Athens. Then I received my MsC diploma in Information Systems and Services-Big Data and Analytics from the Department of Digital Systems, University of Piraeus. For my Masters' thesis "Automated Machine Learning for Clustering Tasks", I designed a schema-agnostic meta-feature extractor based on Kolmogorov-Arnold representation theorem for tabular datasets, using contrastive learning.

Currently, I am a PhD candidate in the Department of Digital Systems, University of Piraeus, working on AutoML for time series, under the supervision of Professor Christos Doulkeridis.

Experience

Skills

• NumPy
• Pandas
• Matplotlib, Seaborn
• PyTorch, TensorFlow
• Mathematical Foundations of Machine Learning
• Mathematical Optimization

Research Interests

My research interests include:

• Machine Learning: Nowadays growth of data availability has led machine learning applications' efficiency and accuracy to an exponential growth. My main interests focuses around Clustering and Time Series Analysis and Forecasting
• Automated Machine Leaning: Machine learning tasks demand data science specialization and domain knowledge, preventing some practitioners from taking part in the ML/AI era. AutoML aims to democratize machine learning, allowing everyone, independently of their field of expertize, to carry out machine learning tasks by automating the task of the machine learning life-cycle. My main interests focuses around AutoML for clustering and time series forecasting.
• Optimization: Optimization is the core concept of AutoML, aiming mainly at the problem of hyperparameter tuning. My main interests focuses around optimization using probabilistic frameworks, such as Bayesian Optimization.

Publication

• Poulakis, Y., Petratos, D., Doulkeridis, C. 2025. PyClust: Building Meta-learning Repositories for Clustering, Internatiol Conference on Data Mining (ICDM 2025) (Just Accepted)
• Emmanouil Dilmperis, Yannis Poulakis, Dimitris Petratos, and Christos Doulkeridis. 2025. Dataset2Graph: A GNN-based Methodology for AutoML for Clustering. In Proceedings of the Workshop on Data Management for End-to-End Machine Learning (DEEM '25). Association for Computing Machinery, New York, NY, USA, Article 11, 1–9. https://doi.org/10.1145/3735654.3735947
• Petratos, D., Poulakis, Y., Pedralba, I.G., Garcia, C.A., Doulkeridis, C. (2025). An AutoML Approach for Bike Demand Forecasting and Redistribution. Intelligent Transport Systems. INTSYS 2024. Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol 608. Springer, Cham. https://doi.org/10.1007/978-3-031-86370-7_9

Activities

• Participated in the 6th ACM Europe Summer School on Data Science, July 2025, Ioannina, Greece
• Presented our paper "An AutoML Approach for Bike Demand Forecasting and Redistribution", 8th EAI International Conference on Intelligent Systems, December 2024, Pisa, Italy
• Participated in the 2nd European Summer School on Artificial Intelligence, July 2024, Athens, Greece
• Participated in the 18th Hellenic Database Management Symposium, July 2024, Athens, Greece
• Participated in the 4th ACM Europe Summer School on Data Science, July 2023, Athens, Greece

Blog