The power of Deep Learning resides in the richness and precision of data we use to train the deep networks. Certainly, now we are at a stage where we have more data than the compute power to utilize the data. Hence, we have come up with strategies where we try to leverage several compute instances in order to learn a model collectively from a larger amount of datasets.

But there is one problem - there exists spatial locality in the availability of the data. Data may reside with different entities, organizations, or companies along with that it may even be…

CS Grad Student at Columbia University, specializing in Machine Learning