ANN4EUROPE

Data processing in experimental high-energy and heavy-ion physics has always been one of the field’s most difficult problems from both a mathematical and computational point of view. Future generation of experiments require the ability to reconstruct and classify data volumes of up to 500 Tbits per second in a computationally affordable and energy efficient way. With the increasing fragmentation of computing architectures and their specialization for AI tasks, successfully tacking this problem will require the development of new algorithms and techniques with AI methods at their heart. As the volume and complexity of the data faced by our experiments exceeds the most complex industrial applications, we can reasonably expect that success in this project will translate to other domains where real-time AI-powered pattern recognition is necessary. Examples include self-driving cars, smart cities, the real-time detection of anomalies in medical procedures or industrial processes, or many others. We will develop deep neural network methods which will allow a cost-effective and energy-efficient processing of data at a variety of experiments in high-energy and heavy-ion physics. Building on top of existing sustainable and open software infrastructures, and our experience with the LHCb (CERN) and CBM (FAIR) experiments in particular, we will make our advances available to all who wish to benefit from them. We have decades-long and highly complementary experience in developing fast, reliable and highly efficient real-time processing, as well as using artificial neural networks at all the most time-consuming stages of experiments. We equally have a proven track record of developing original data processing solutions across a range of computing architectures and taking these original ideas from R&D projects to production. A collaboration among our teams will allow us to use AI in order to maximally leverage the power of tomorrow’s hybrid computing architectures, delivering orders-of-magnitude improvements in computational and energy efficiency and powering the fundamental research of tomorrow.