Alexander Sokol at CompatibL explains how trading and risk models are getting an exponential boost from machine learning.
For this podcast, Alexander Sokol, founder, executive chairman, and head of quant research at CompatibL Technologies, focuses on how trading and risk models are getting an exponential boost from machine learning.
For instance, the standard models for interest rates are based on incremental changes in rates but those models are falling short now as rates are in turmoil across the globe as central bankers raise rates to drive down inflation.
Alexander Sokol
“So, the new model types that we developed, which are called Autoencoder Market Models, are based on training the models to the entire history of interest rates not only for the currency we are modeling but also across all other currencies. This historical data, that we are training our model to, represents all kinds of market regimes,” Sokol says. “So, by training the model to all of the rates regimes across all currencies, these models become more effective following the change in market regime.”
Essentially, the models using machine learning can look back in time to find conditions that could be similar to current conditions. By taking that history into consideration, the new models can be more accurate.
The interesting fact is that this is just the start, Sokol says. Once financial services firms have laid the groundwork through cloud computing and other forms of modernization, then firms can start to explore variational autoencoders and other innovations at a time when they might need them the most.
The podcast also covers:
- A definition of a variational autoencoder (VAE) and CompatibL’s Autoencoder Market Models (AEMM);
- How machine learning-enabled models might help with interest rate portfolios, managing limits and add-ons, and credit exposure concerns; and
- How cloud computing and the push for digital transformations are helping to advance autoencoder-based models.
CompatibL is a trading and risk solutions provider that Sokol founded in 2003.
In 2022, Alexander was voted FinTech Person of the Year via the FTF Awards competition. CompatibL also won the Best Digital Transformation Solutions Provider award for 2021.
Need a Reprint?
Leave a Reply