Blended Learning-Assimilating Authentic Data Into Deep Learning Models

Abstract

The age of deep learning is picking up in a way that increases the curiosity in man to make the world of predictions as realistic as possible. In pursuit of achieving this goal, he comes up with approximate algorithms[8], that predict satisfactorily with long training despite the use of GPUs. A deep learning model is not perfect, unless it accommodates new trends and the data of latest discovery that would impact significantly in future inferences. Assimilating this critical data into the pre-trained deep learning model[6]involves validity of the data. A Blockchain-like structure could be incorporated atop the data, validate and introduce the authentic data to the under-performing pre-trained model. We call this a blended learning which uses blockchainified data to fine tune the model. This idea of secure data assistance to update pre-trained model opens up a novel field of research that brings a synergy between Artificial Intelligence(AI) and Blockchain.