A decentralized aggregation mechanism for training deep learning models\n using smart contract system for bank loan prediction

Abstract

Data privacy and sharing has always been a critical issue when trying to\nbuild complex deep learning-based systems to model data. Facilitation of a\ndecentralized approach that could take benefit from data across multiple nodes\nwhile not needing to merge their data contents physically has been an area of\nactive research. In this paper, we present a solution to benefit from a\ndistributed data setup in the case of training deep learning architectures by\nmaking use of a smart contract system. Specifically, we propose a mechanism\nthat aggregates together the intermediate representations obtained from local\nANN models over a blockchain. Training of local models takes place on their\nrespective data. The intermediate representations derived from them, when\ncombined and trained together on the host node, helps to get a more accurate\nsystem. While federated learning primarily deals with the same features of data\nwhere the number of samples being distributed on multiple nodes, here we are\ndealing with the same number of samples but with their features being\ndistributed on multiple nodes. We consider the task of bank loan prediction\nwherein the personal details of an individual and their bank-specific details\nmay not be available at the same place. Our aggregation mechanism helps to\ntrain a model on such existing distributed data without having to share and\nconcatenate together the actual data values. The obtained performance, which is\nbetter than that of individual nodes, and is at par with that of a centralized\ndata setup makes a strong case for extending our technique across other\narchitectures and tasks. The solution finds its application in organizations\nthat want to train deep learning models on vertically partitioned data.\n