On 3 June 2015 the Royal Statistical Society's Applied Probability section held a meeting on stochastic approximation at Errol Street.
Pierre Tarres (Paris-Dauphine) spoke on language dynamics and social networks. This talk considered the 'signalling game', where two players have a set of signals and a set of meanings and use reinforcement to develop a relationship between them, ie a language, and a related social network game, using stochastic approximation theory to investigate the possible limiting behaviour.
Teo Sharia (Royal Holloway) spoke on stochasic approximation and online estimation algorithms, analysing online algorithms for maximum likelihood and related estimators using stochastic approximation methods.
Vladislav Tadic (Bristol) spoke on stochastic gradient search, with two sections, one on convergence rate in stochastic gradient search, and one on asymptotic behaviour of stochastic gradient search with biased gradient estimation. Both sections introduced existing results which have some restrictions, and then introduced new results.
Finally, Steven Perkins spoke on abstract stochastic approximation for learning in games, the aim being to generalise convergence results to games with continuous action sets by considering stochastic approximation on Banach spaces.