Machine Learning and High Performance Computing

Sapiat believes that advances in Machine Learning techniques coupled with the rise of contemporary cloud-based cluster computing environments, massive parallel architectures offer new possibilities for Analytics in finance. These, combined with new and sometimes exotic data sources allow for traditionally difficult questions to be addressed.

Combining a structural economic approach with high frequency macroeconomic data allow for longer-horizon analyses. Relational learning techniques allow for investors to look for consistency across forecasts from different sources. Non-linear approaches to pattern recognition permit us to identify new regimes and cross-movements across assets in a variety of market conditions.

Longer-horizon, multi-asset class institutional investors have long had the problem of making risk and return forecasts where uncertainty, estimation error, and human bias have dominated any particular model’s ability to provide precision. We believe that this can be addressed with the judicious use of expert knowledge to finely calibrate the questions and approaches that are most usefully elucidated by machine learning.