Incremental Machine Learning with Holographic Neural Theory for ATD/ATR

Machine learning has been used intensively since the past 30 years to discriminate pixels from background or objects of interest from other classes of objects by training on a set of relevant features. As image sources are now producing more images that we can realistically cope with, the goal is to explore the limits of these approaches for ATD/ATR in order to optimally define the domains in which decisions can be left to automated processes or should require human intervention. With this objective in mind, this paper presents an assessment of the performances of the holographic neural technology (AND Corporation) to support applications that would require incremental learning