How Do Machines Learn?

“Children don’t have adults telling them what each pixel represents in every image they see, or what are the objects present in every image, what is the grammatical structure and the fine sense of every word in every sentence they hear. We extract most of the information from simple observation, and that is what unsupervised learning in principle does.”

Sometimes researchers combine these approaches in a method called “semi-supervised learning.” In this approach, machine learning algorithms are given a small amount of labeled training data and a much larger pool of unlabeled data from which to learn. This approach can combine the best of both worlds—improved accuracy associated with supervised machine learning and the ability to make use of unlabeled data, as in the case of unsupervised machine learning.

Interested in learning more?

Read the Machine Intelligence Primer.

1 - 4 of 8