Neural-network solutions to stochastic reaction networks

Efficient Fisher Information Computation and Policy Search in Sampled Stochastic Chemical Reaction Networks through Deep Learning

“…The emergence of an expanding number of Neural Network models designed to approximate the mass function of Reaction Networks holds the possibility of enhancing the accuracy of the methodology presented here in the future. Most notably, the Variational Autoregressive Network architecture has been demonstrated to successfully estimate joint distributions, even in high-dimensional settings [8]. Still, at present, the parameters of the Reaction Network are not explicit inputs to the surrogate model, which means sensitivities cannot be obtained in the same way as in our approach.…”

“…We also demonstrate how this Neural Network can be used to perform fast policy search. The efficiency of these approaches is illustrated on a set of examples, and is compared to that of the current state-of-the-art.Recently developed Neural Networks offer a new way of computing the probability mass function of Chemical Reaction Networks [6,7,8]. As in many other fields, the introduction of Neural Networks has been motivated by their universal approximation property, their generalisation ability and their compositional structure [9].…”

“…Recently developed Neural Networks offer a new way of computing the probability mass function of Chemical Reaction Networks [6,7,8]. As in many other fields, the introduction of Neural Networks has been motivated by their universal approximation property, their generalisation ability and their compositional structure [9].…”