New project started
We started a new research project CogXAI.
"Cognitive neuroscience Inspired techniques for eXplainable AI" (CogXAI)
Training Artificial Neural Networks (ANNs), known as Deep Learning (DL), has become a very popular tool for solving challenging tasks across various fields of application. Performance gains are often achieved through increasing their complexity, which complicates explaining the learned models and makes their desicions more opaque. ANNs are inspired by the human brain, which is known as one of the most complex systems. Over the past 100 years, neuroscientists developed a variety of sophisticated tools to study the real brain's structure, function and the associated human behavior. This wealth of experience is hardly getting any attention in DL.
In CogXAI, we will research how use this potential to improve transparency and explainability of ANNs. We will investigate, how methods and findings from cognitive neuroscience can be adapted for ANNs. This will involve both innovative analysis techniques for ANNs and novel network architectures, which are per-design better interpretable.