Exploring Unsupervised Deep Learning
Unsupervised learning works with data that does not have any labels. More broadly, unsupervised learning aims to uncover the intrinsic patterns hidden within the data. The most rigorous and expensive part of a supervised machine learning project is the labels required for a given data. In the real world, there is tons of unlabeled data available with tons of information that could be learned from. Frankly, it’s impossible to obtain labels for all of the data that exist in the world. Unsupervised learning is the key to unlocking the potential of the abundant unlabeled digital data we have today. Let’s explore a hypothetical situation below to understand this better.
Imagine that it costs 1 USD and 1 minute to obtain a label for a row of data for whatever use case it could be, and a single unit of information can be obtained through supervised learning. To get 10,000 units of information, 10,000 USD would need to be spent, and 10...
