Abstract

We propose to use a special type of generative neural networks – a Restricted Boltzmann Machine (RBM) – to build a powerful generator of synthetic market data that can replicate the probability distribution of the original market data. An RBM constructed with stochastic binary activation units in both the hidden and the visible layers (Bernoulli RBM) can learn complex dependence structures while avoiding overfitting. We consider an efficient data transformation and sampling approach that allows us to operate Bernoulli RBM on real-valued data sets and control the degree of autocorrelation and non-stationarity in the generated time series.

Speaker bio

In his role as Managing Director and Global Head of Data Analytics, Alexei is responsible for providing data analytics services to Corporate, Commercial and Institutional Banking division of Standard Chartered Bank.

He joined Standard Chartered Bank in 2010 from Barclays Capital where he managed a model development team within Credit Risk Analytics. Prior to joining Barclays Capital in 2004, he was a senior quantitative analyst at Dresdner Bank in Frankfurt.

Alexei holds MSc in Theoretical Nuclear Physics from the University of Kiev and PhD in Mathematical Physics from the Institute for Mathematics, National Academy of Sciences of Ukraine.

He was the recipient of the 2019 Quant of the Year award from Risk magazine.