The new ATLAS Fast Calorimeter Simulation

The physics and performance studies of the ATLAS detector at the Large Hadron Collider re- quire a large number of simulated events. A GEANT4 based detailed simulation of the ATLAS calorimeter systems is highly CPU intensive and such resolution is often unnecessary. To reduce the calorimeter simulation time by a few orders of magnitude, fast simulation tools have been developed. The Fast Calorimeter Simulation (FastCaloSim) provides a parameterised simulation of the particle energy response at the calorimeter read-out cell level. In Run 1, about 13 billion events were simulated in ATLAS, out of which 50% were produced using fast simulation. For Run 2, a new parameterisation is being developed to improve the original version: it incorporates developments in geometry and physics lists during the last five years and benefits from the knowledge acquired from the Run 1 data. The algorithm uses machine learning techniques to improve the parameterisations and to optimise the amount of information to be stored in the ATLAS simulation infrastructure. The latest developments of the new ATLAS FastCaloSim are reviewed in this proceeding.