Unsupervised learning

In supervised learning, we have features $x_i$ and class labels $y_i$. Write a program that produces $y_i$ form $x_i$

In unsupervised learning, we only have $x_i$ values, but no explicit target labels. We can

• Outlier detection: is this a normal $x_i$?
• Similarity/Clustering: which examples look like this $x_i$
• Which $x_i$ occur together
• Latent-factors: what ‘parts’ are the $x_i$ made from
• Data visualization: what does the high-dimension $X$ feature space look like?
• Ranking: what are the most important $x_i$?