Keynote Talks for ACSS 2027 will be Announced Soon.

Previous Year's Keynote Talks (ACSS 2026)

Prof. Sanghamitra Bandyopadhyay

Indian Statistical Institute, Kolkata, India



Artificial Intelligence and Machine Learning Algorithms in Life Sciences

Artificial Intelligence (AI) and Machine Learning (ML) algorithms find extensive applications in all domains of scientific research, in particular in diverse areas of life sciences for making novel discoveries and for gaining deeper insights into various processes of life. In this talk we will first present brief overviews of artificial intelligence, machine learning and molecular biology. In particular, a quick introduction to the central dogma of molecular biology will be provided, which is fundamental to the understanding of a large class of machine learning applications in biology. Various kinds of data sets emerging in different areas of healthcare research will be mentioned. We will then discuss some case studies in a few areas of biology, namely, molecular target prediction, a graph theoretic method for biomarker identification and graph based approaches in drug interaction studies. The talk will conclude with a mention of some issues and challenges in this area.

Prof. Piotr A. Kowalski

AGH University of Krakow, Poland



Sensitivity Analysis as an XAI Method for Neural Network Analysis and Pruning

As neural networks are increasingly deployed in complex real-world applications, the need to understand, justify, and improve their decision-making processes becomes ever more important. This invited talk will focus on Sensitivity Analysis as a powerful method within Explainable Artificial Intelligence (XAI), offering insights into how neural network outputs respond to changes in inputs, parameters, neurons, or the model's structure. The presentation will introduce the fundamental idea of sensitivity-based explanations: identifying which features, weights, neurons, or layers have the strongest influence on the model's behaviour. By analysing a neural network's response to controlled perturbations, gradient-based measures, or other sensitivity indicators, it is possible to reveal hidden dependencies, detect fragile or redundant components, and better understand the internal mechanisms of black-box models. Particular attention will be devoted to the role of Sensitivity Analysis in neural network pruning. The talk will show how sensitivity measures can guide the removal of less important weights, neurons, filters, or connections, leading to more compact and efficient models while preserving predictive performance. In this context, sensitivity-based pruning will be discussed not only as an optimisation technique, but also as an explainability-driven approach that helps expose which parts of the network are truly essential to its decisions. Through illustrative examples, the presentation will demonstrate how Sensitivity Analysis supports model interpretation, validation, compression, and trustworthiness. Its advantages, limitations, and its relation to other XAI methods, such as saliency maps, feature-importance techniques, LIME, and SHAP, will also be considered. The talk will highlight the importance of combining explainability with efficiency, showing how XAI methods can contribute both to human understanding and the design of more transparent, reliable, and resource-aware neural network systems.