UNIVERSITY OF HERTFORDSHIRE
             COMPUTER SCIENCE RESEARCH COLLOQUIUM

                            presents

     
	"Prototype-based learning for decision making and segmentation
         -- steps and problems towards explaining decisions"
	

                    Prof. Lehel Csato
              (Faculty of Mathematics and Computer Science,
            Babeș-Bolyai University, Cluj-Napoca, Romania)

                        12 November 2025
                         13:00 -14:00 
                          Room LB216

                  Everyone is Welcome to Attend
				  Refreshments will be provided
				  
				                               

Abstract:
Neural network models are being used with great success, yet for critical systems the measurements in terms of accuracy or error rates might be insufficient; there is need for the decisions to be grounded. A fairly important research direction within the neural network research is to make these models as explainable - meaning that the decisions are grounded - as possible. A relatively new explainable research direction is the consideration of prototypes being implemented within the neural system.

We detail our experiments and our results in applying prototype-based architectures for classification and segmentation. One meets several critical decisions in implementing the system: (a) what type of backbone to use, and (b) what additional criteria to optimize. We present experimental results, draw conclusions, and highlight some future research directions.

Biography:
Lehel Csató is a professor within the Faculty of Mathematics and Computer Science at the Babeș-Bolyai University in Cluj-Napoca, Romania. He had finished his PhD  at Aston University (UK) and the focus of his research was the usage of probabilistic non-parametric methods for inference using latent variable models. He is interested in mathematical aspects of machine learning, providing approximate solutions for inverse problems, performing sparse inference on large systems with applications to robotics, classification of complex data, and active learning. With the onset of the deep learning models, his interests shifted to an understanding of the working of deep systems; his research includes the mathematical understanding of self-explainable deep learning models and model simplification using pruning.

---------------------------------------------------
Hertfordshire Computer Science Research Colloquium
http://cs-colloq.cs.herts.ac.uk