The two DTU Compute researchers Postdoc Trine Julie Abrahamsen and Professor Lars Kai Hansen have discovered that computers making decisions based on big data often have large so-called ’blind spots’.
The blind spots may arise when the computer combines many and uncertain factors in new data. In the worst-case scenario the computer may overlook important information and draw the wrong conclusion.
Mathematical trick
Trine Julie Abrahamsen and Lars Kai Hansen’s solution is based on a mathematical trick which may help to discover and compensate for any blind spots. This enables the computer to act more correctly.
This will have a great impact on a number of fields from medical engineering to e-commerce.
Trine Julie Abrahamsen and Lars Kai Hansen have further developed the so-called ’support vector machine’ invented by the Russian mathematician Vladimir Vapnik and his colleagues in 1992.
The support vector method is applied to diagnose diseases on the basis of scan images and genetics, in bank credit ratings of the customers and also automatic document analysis and online advertising.