Search for: All records

Chaotic dynamics are ubiquitous in many real-world systems, ranging from biological and industrial processes to climate dynamics and the spread of viruses. These systems are characterized by high sensitivity to initial conditions, making it challenging to predict their future behavior confidently. In this study, we propose a novel deep-learning framework that addresses this challenge by directly exploiting the long-term compounding of local prediction errors during model training, aiming to extend the time horizon for reliable predictions of chaotic systems. Our approach observes the future trajectories of initial errors at a time horizon, modeling the evolution of the loss to that point through the use of two major components: 1) a recurrent architecture (Error Trajectory Tracing) designed to trace the trajectories of predictive errors through phase space, and 2) a training regime, Horizon Forcing, that pushes the model’s focus out to a predetermined time horizon. We validate our method on three classic chaotic systems and six real-world time series prediction tasks with chaotic characteristics. The results show that our approach outperforms the state-of-the-art methods. 

Implantable tubes, shunts, and other medical conduits are crucial for treating a wide range of conditions from ears and eyes to brain and liver but often impose serious risks of device infection, obstruction, migration, unreliable function, and tissue damage. Efforts to alleviate these complications remain at an impasse because of fundamentally conflicting design requirements: Millimeter-scale size is required to minimize invasiveness but exacerbates occlusion and malfunction. Here, we present a rational design strategy that reconciles these trade-offs in an implantable tube that is even smaller than the current standard of care. Using tympanostomy tubes (ear tubes) as an exemplary case, we developed an iterative screening algorithm and show how unique curved lumen geometries of the liquid-infused conduit can be designed to co-optimize drug delivery, effusion drainage, water resistance, and biocontamination/ingrowth prevention in a single subcapillary–length-scale device. Through extensive in vitro studies, we demonstrate that the engineered tubes enabled selective uni- and bidirectional fluid transport; nearly eliminated adhesion and growth of common pathogenic bacteria, blood, and cells; and prevented tissue ingrowth. The engineered tubes also enabled complete eardrum healing and hearing preservation and exhibited more efficient and rapid antibiotic delivery to the middle ear in healthy chinchillas compared with current tympanostomy tubes, without resulting in ototoxicity at up to 24 weeks. The design principle and optimization algorithm presented here may enable tubes to be customized for a wide range of patient needs.