Deep symbolic regression

Using a suitable reward mechanism with the reinforcement learning to identify the correct mathematical expression from an exponentially growing space of expressions that may describe a given dataset.

Long-term prediction of chaotic systems

A new deep recurrent architecture capable of learning the state evolution of various chaotic dynamical systems, substantially extending the prediction horizon.

Interpretable Spatio-Temporal Modeling

Combined ensemble learning and the multi-Markov-Blankets concept in Bayesian probability theory to provide accurate predictions of extreme precipitation events.

Crime Hot Spot Forecasting

Integrated transposed convolutions and deep recurrent networks to capture complex spatio-temporal patterns in crime data and generate accurate forecasts of crime hotspots.