Deep Learning is picking momentum in Quantitative Finance, outside the apparent application to the prediction of asset costs (where to my understanding it is not particularly effective) and spreading into the more severe application area of option rates and threat management.
These 2 recent papers clearly show the advantages of DL as a pricing innovation option to the classical FDM and Monte-Carlo in specific contexts:
These posts likewise check out the ideas of ‘deep calibration’ (application of DL to the calibration of financial designs in a particularly intriguing kind) and ‘deep analytics’ (utilizing DL to solve the conundrum of revaluation in the context of regulatory simulations):
I must include that those two posts only constitute early expeditions. Those are extremely interesting and high prospective ideas, but serious implementation obstacles should be gotten rid of to put them in production.