The distinction between "supervised" and "unsupervised" learning is fundamentally superficial -- an artifact of the semantics of the last couple decades of ML. All learning is guided by an objective (including all forms of ML and all forms of human learning).
-
Show this thread
-
What matters to progress towards greater generality in AI is *how* you learn -- for instance, gradient descent with parametric functions vs discrete search over a discrete program space. The supervision objective is an implementation detail. There will always be one anyway.
1 reply 7 retweets 109 likesShow this thread -
Now, from the standpoint of doing applied ML, the ability to solve problems without human labeling effort is huge. So unsupervised learning is a practical concept, but one that is theoretically irrelevant. It affects how you use ML, while being orthogonal to ML algorithms.
1 reply 8 retweets 95 likesShow this thread -
The important questions, from the applied ML angle, are: 1. What problems can you solve using data reconstruction as your objective? 2. What automatically-generated or passively-collected alternative to human-provided labels can you use to solve task X?
1 reply 22 retweets 138 likesShow this thread
These are broadly "data science" questions as opposed to algorithmic questions.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.