Giacomo Ziffer and Federico Giannini, Ph.D. students in Information Engineering under the supervision of Prof. Emanuele Della Valle from the Department of Electronics, Information, and Bioengineering at Politecnico di Milano, won the Best Student Runner-Up Paper Award at the IEEE Big Data 2024 international conference, held in Washington, D.C., from December 15 to 18, 2024.

Their paper, titled “Tenet: Benchmarking Data Stream Classifiers in the Presence of Temporal Dependence," earned recognition in a highly competitive setting. The conference attracted 661 submissions from 2,505 authors across 53 countries, with an overall acceptance rate of just 18%.

The article examines the limitations of the traditional assumption of independent and identically distributed (i.i.d.) samples and highlights how failing to account for temporal dependencies in data streams can negatively affect the performance and reliability of Streaming Machine Learning (SML) models. In particular, it emphasizes the risk of errors in both the design and evaluation of these models.

To address this, the paper introduces Tenet, a benchmarking framework for evaluating data stream classifiers in non-i.i.d. scenarios. Tenet includes a data stream generator that introduces temporal dependencies and a baseline continuous Long Short-Term Memory algorithm (cLSTM). Experimental results show that cLSTM outperforms state-of-the-art SML classifiers in time-dependent data streams, emphasizing the need for further research in this area and establishing Tenet as the first public benchmark for such studies.