Optimal posting of collateral with recurrent neural networks

Pierre Henry-Labordère applies neural networks to a control problem approach for managing collateral

CLICK HERE TO VIEW THE PDF

In this article, Pierre Henry-Labordère considers the problem of optimal posting of collateral in two assets. The associated stochastic control problem is then solved by parameterising the optimal control with a (recurrent) neural network (NN). Then, the training of the NN is achieved with a backpropagation algorithm, possibly complemented by a particle swarm optimisation (or simulated annealing)

Since the recent financial crisis, collateralisation has become a