How machines learn?

By | March 27, 2017

Writing in Datanami, a news portal dedicated to Big Data news, insights and analysis, Fiona McNeill, a SAS global marketer and Dr. Hui Li,  senior staff scientist for SAS, shares light on how machines learn. Starting with examples of various enterprises using machine learning to design personalized offerings to attract customers, they then raise the important point of different vendors jumping on the bandwagon of machine learning with their own approaches and solutions making the whole thing confounding to the user. And thru’ this article in Datanami, the duo from SAS try to unravel machine learning and make it easier for users to understand how exactly machine learning works.

Machine learning models are designed to learn how to perform tasks and with algorithms designed to see relationships and patterns between various factors, these models learn continuously from data.  And to generalize this model for business they are then validated on whole set of new data not used initially for training. These models may be made to learn in different ways like supervised learning, semi-supervised, unsupervised and reinforcement learning.

Machine learning is at the center of many advanced intelligent solutions emerging now, like AI, Neural networks, Natural Language Process and Cognitive Computing.

  • Artificial Intelligence – A discipline enabling the design of machine with problem solving skills to accomplish tasks just as human beings can.
  • Neural Networks and Deep Learning – Neural networks are programs written to learn from observational data and present solution to the problem on hand. These are used in speech and image recognition and are very successful in supervised learning.
  • Natural Language Processing and Cognitive Computing – NLPs are interfaces which enable machines to understand human language and humans to interpret machine output. These are applied in image captioning, text generation and machine translation.

The confluence of Big Data and massive parallel computational environments are driving the machine learning initiatives and the goal is to deliver solutions that can be highly customizable and with human – like cognition features.

For more on this, please visit:

Leave a Reply

Your email address will not be published. Required fields are marked *