Dan Zitting serves as Chief Product & Strategy Officer at Galvanize, the global leader for GRC software. Current economic challenges and the ongoing public health crisis have transformed the circumstances in which fraud happens. The good news is that the tools to address it are at the ready. Machine learning gives organizations the ability to fight both internal and external fraud threats to reduce risk. Regardless of global conditions, there are a few basic elements that fuel fraud.

Delhi: In one of the biggest exposes of 2016, it has been revealed that in order to bag tender for biometric facial recognition device, Cryptometrics, a company in Canada, gave kickbacks to officers of Air India. Biometric facial recognition device is used for recognition of faces of passengers. In an exclusive report, Zee Media Corp has learnt that on 24 February 2006, Air India had issued a Request for Proposal for the device and the tender for the same given by 20 companies including Canada's Cryptometrics. Later, it was revealed in order to get the tender, Cryptometrics paid kickbacks to Air India officials through a person named Nazir Karigar. His closeness with Air India officials can be ascertained from the fact that Nazir had the full copy of the tender with him on 28 December 2005 itself. Whereas Air India had issued the tender on 24 February 2006.

Error backpropagation is an extremely effective algorithm for assigning credit in artificial neural networks. However, weight updates under Backprop depend on lengthy recursive computations and require separate output and error messages — features not shared by biological neurons, that are perhaps unnecessary. In this paper, we revisit Backprop and the credit assignment problem. We first decompose Backprop into a collection of interacting learning algorithms; provide regret bounds on the performance of these sub-algorithms; and factorize Backprop's error signals. Using these results, we derive a new credit assignment algorithm for nonparametric regression, Kickback, that is significantly simpler than Backprop. Finally, we provide a sufficient condition for Kickback to follow error gradients, and show that Kickback matches Backprop's performance on real-world regression benchmarks.