CAIMAN: Continual Artificial Intelligence Models for Atmospheric Convection in Subtropical Regions
A one year long project, started in November 2023, where the central guiding question is: “the development of climate models that can continuously adapt to and monitor the different aspects of atmospheric convection to predict severe weather conditions.
Atmospheric convection is an important component of climate modeling, because of its role in the distribution of energy and trace gases (e.g., water vapor, carbon monoxide, and aerosols). Strong (deep) convection can cause severe weather conditions such as thunderstorms, hail, heavy rain, and flooding. An accurate representation of atmospheric convection in global and regional climate models is required to predict the latter. However, current dynamical models face a number of ongoing challenges. First, without parametrization, resolving convection at high resolution can be computationally prohibitive, whereas coarser resolutions cannot explicitly account for convection and thus produce unacceptable forecasts. Second, and perhaps more importantly, complex interactions between mesoscale, orographic, and local phenomena contribute to the rapid development of various modes of convection, particularly in subtropical regions like the Indian subcontinent. Classifying these accurately determines whether havoc in one region (e.g. several Himalayan flood cases) can be anticipated correctly while representing precipitation properly in other regions. Finally, because the climate is changing over time, continuous adaptation to the temporal evolution of the factors that cause different types of convection is required. Project CAIMAN: Continual Artificial Intelligence Models for Atmospheric Convection in Subtropical Regions, will catalyze an innovative interdisciplinary solution to above challenges. As the key novelty, CAIMAN will kick-start AI supported identification of convection processes, centered in cutting-edge continual learning techniques that enables updates over long periods of time and adapting to climate changes without having to permanently store data. As a case study, we will investigate the product of convection, i.e. cloud feature development, leveraging long-term satellite data and simulation over multiple years on an hourly to daily scale.
SyCLeC: Symposium on Continual Learning beyond Classification
A three day in-person symposium from Nov 27th to Nov 29th in Mainz, Germany.
Continual, or lifelong, machine learning addresses the crucial questions that arise when aiming to overcome the limitations of single training cycles, rigid inference engines, and fixed datasets. At the center, it acknowledges that data selection, models, training algorithms, and evaluation measures are not static. Despite its recently emerged popularity, current research has just started to grasp how we can accommodate these factors into human-like lifelong learning AI systems. Although presently ongoing efforts start to take into account complex sequences of datasets, they focus predominantly on image classification tasks. Unfortunately, such tasks significantly simplify learning, e.g. by only performing image-level recognition, assuming that all data is always labelled and that continual learning consists mainly of learning to recognize new object types. In this initiative, we thus set out to bring together researchers who are at the forefront of continual learning and computer vision to jointly catalyze the foundation for continual learning beyond classification. In an inaugural symposium, we will establish common grounds and discover synergies towards continual learning for a plethora of relevant computer vision tasks, such as semantic segmentation and learning that does not require labels.