Let's Flow within Kubeflow - Intel AI

In this blog post, we will go through how to train MNIST using distributed Tensorflow* and Kubeflow* from scratch. Machine learning (ML) and deep learning (DL) have been around for more than half a century now, yet it is just as of late that these ideas have begun to flourish--thanks to advancements in compute capabilities and the deluge of data. This is due, essentially, to the fact that ML/DL algorithms need vast amounts of information to register the desired level of accuracy. Likewise, this high volume of data requires high processing power so it can yield the expected intelligence and knowledge. With the emergence of Cloud and other distributed frameworks, we started to treat a set number of servers as "cattle versus pets" in an attempt to utilize their collective assets for storage and computation.