Located at the Jagiellonian University in Kraków, Poland, our research group delves into the frontier of artificial intelligence, striving for interpretable and sustainable solutions.

Overview: The modern era has experienced rapid advancements in deep learning methods, making significant strides in various sectors. However, the predominant ‘black-box’ nature of deep neural networks poses challenges that can lead to severe societal impacts affecting health, freedom, social bias, and safety.

Our Approach: While the current landscape offers many post hoc explainability methods, our group believes in replacing black-box models with inherently interpretable ones. Drawing inspiration from human problem-solving strategies, we’re particularly interested in modern case-based reasoning in deep computer vision, such as the Prototypical Part Network (ProtoPNet). This approach compares encoded input images to learned prototypical parts to generate intuitive and faithful explanations.

Our Goals: Our primary objective is to harness the potential of the prototypical approach, making it more intuitive and sustainable. We believe that these improved methods can be applied in open-world settings, like self-supervised learning or continual learning.

Our research is spread across four pivotal tasks:

• Establishing an evaluation pipeline that measures model accuracy and interpretability
• Developing sustainable and interpretable deep learning models
• Providing precise and intuitive explanations
• Applying novel case-based reasoning models for life sciences.

Our pioneering approach is backed by extensive research and a proven track record in the domain of interpretability and life science applications of machine learning. With experience in leading cutting-edge research projects in computer vision, we are positioned at the forefront of creating a trustworthy, society-oriented AI.

Join us in our mission to revolutionize the realm of AI, making it more interpretable, sustainable, and intuitive for a better future.