#Machine learning#fintech#AI#Banking

Machine learning gives Danske transformative quant boost

Danske Bank combines AI with algorithmic patterns to rapidly process quant analysis

|Oct 5|magazine4 min read

In 2006, a new mathematical technique was introduced to banking. The adjoint algorithmic differentiation (AAD) allowed quantitative analysts (or quants) to run modelling that produced accurate simulation results at high speeds. The development shook finance and soon derivatives markets were being commonly navigated by algorithmic calculations of credit value adjustments (CVAs).

Now, two Copenhagen-based quants, Brian Huge and Antoine Savine, have improved the technique by twinning it with machine learning that spots patterns in the datasets. Building their neural networks on top of the existing quant architecture, the pair think they can produce comparable results thousands of times faster.

“The pathwise differentials in themselves contain plenty of information and we suspected we could use it to train pricing approximations more effectively,” Huge told Risk . “It can also be used to generate dynamic risk reports, because we can quickly compute risks in a vast number of future scenarios.”

“Everything we can do with the differential machine learning can also be done already with nested Monte Carlo simulation,” says Savine. “With differential machine learning you just do it thousands of times faster and with similar accuracy.”

Giuseppe Benedetti, senior quantitative analyst at financial software vendor FIS, says: “Every time you increase speed, the quality of your risk management massively improves, because when calculations are slow, people have to cut corners.”

He adds: “We’re looking into this technique as a candidate to be implemented in our products. Using AAD pathwise derivatives to regularise the exposure estimation process is a very clever idea to reduce noise and make the methodology more reliable and applicable in production environments.”

Danske Bank is testing the combined model for market risk before implementing it on a wider scale. It is expected to be rolled out in 2021.

Read more
Read more