Bank of America Corporation
SYSTEM FOR INTELLIGENT DRIFT MATCHING FOR UNSTRUCTURED DATA IN A MACHINE LEARNING ENVIRONMENT

Abstract:

Systems, computer program products, and methods are described herein for intelligent drift matching for unstructured data in a machine learning environment. The present invention is configured to receive an original unseen image from a computing device; initiate a convolutional encoder-decoder algorithm on the original unseen image; generate a reconstructed unseen image; generate an unseen input dataset based on at least the reconstructed unseen image; determine a reconstruction loss associated with the original unseen image; determine that the reconstruction loss associated with the original unseen image matches a first reconstruction loss associated with a first training image, wherein the first training image is associated with one or more training images; retrieve, from the data repository, a first set of parameters associated with the first training image based on at least determining the match; classify, using the first set of parameters, the original unseen image into one or more class labels.