The accuracy of simulation-based forecasting in chaotic systems is heavily dependent on high-quality estimates of the system state at the beginning of the forecast. Data assimilation methods are used to infer these initial conditions by systematically combining noisy, incomplete observations and numerical models of system dynamics to produce highly effective estimation schemes. We introduce a self-supervised framework, which we call \textit{amortized assimilation}, for learning to assimilate in dynamical systems. Amortized assimilation combines deep learning-based denoising with differentiable simulation, using independent neural networks to assimilate specific observation types while connecting the gradient flow between these sub-tasks with differentiable simulation and shared recurrent memory. This hybrid architecture admits a self-supervised training objective which is minimized by an unbiased estimator of the true system state even in the presence of only noisy training data. Numerical experiments across several chaotic benchmark systems highlight the improved effectiveness of our approach compared to widely-used data assimilation methods.

The accuracy of simulation-based forecasting in chaotic systems is heavily dependent on high-quality estimates of the system state at the beginning of the forecast. Data assimilation methods are used to infer these initial conditions by systematically combining noisy, incomplete observations and numerical models of system dynamics to produce highly effective estimation schemes. We introduce a self-supervised framework, which we call \textit{amortized assimilation}, for learning to assimilate in dynamical systems. Amortized assimilation combines deep learning-based denoising with differentiable simulation, using independent neural networks to assimilate specific observation types while connecting the gradient flow between these sub-tasks with differentiable simulation and shared recurrent memory. This hybrid architecture admits a self-supervised training objective which is minimized by an unbiased estimator of the true system state even in the presence of only noisy training data.

It manages our phones and homes, helps us navigate, and advises us what to watch, read, listen to, and buy. Soon it will transform our health, says trauma surgeon and data-science expert Rachael Callcut, MD, MSPH. There is a certain amount of bias that we, as humans, bring to any clinical scenario: Without even realizing it, we may look past critical pieces of information that could help our patients get better. AI, which is essentially a computer algorithm that learns from data, can uncover patterns that we can't see – either because of those biases or because the human brain simply can't assimilate the vast quantity of medical data that is now available from hospital sensors and other digital health devices. Ultimately, AI promises to reduce human error and make our care more efficient, which will improve outcomes for our patients.

The smart city challenges are not a challenge for AI…but for us'. From the day one, human civilisation has always tried to seek out ways that could make our life better and better each day by overcoming the challenges that come by. We always look for new ideas, innovations, and strategies that could augment our existence as effectively as possible – as they say, the sky's the limit. And even with artificial intelligence, it's the same – smart city challenges are easy for AI to be accomplished but how it does accomplish is also important. The path chosen to reach the destination is more important than the destination itself.