In unsupervised learning, the goal is to infer structure or patterns from data without needing any prior labeling of the data. Because the data is unlabeled, there is typically no way to evaluate the accuracy of the learning algorithm, a major distinction from supervised learning. Unsupervised learning algorithms typically are not given any a priori knowledge of the data, except perhaps indirectly by the tuning parameters given to the algorithm itself.
Unsupervised learning is commonly used for problems that might be solvable by eye if the data had very few dimensions, but the large dimensionality of the data makes this impossible or very difficult for a human to infer. Unsupervised learning can also be used for lower-dimension ...
