/* ---- Google Analytics Code Below */

Wednesday, August 26, 2015

Mysteries, Experiences and Value of Unsupervised Learning

I am reminded of  a conversation with a manager  early in my data analytics career when I mentioned using "unsupervised learning" ... .  He asked in essence:   "How can you find a result if you don't supervise the process?   It sounds so messy and casual"  I answered:   " We will first simply observe and learn, so we can later  supervise ... ".    You can start with just observation.  Just like you should start every project with visualizing your data.

In O'Reilly. a David Beyer interview with " .... Ilya Sutskever ... a research scientist at Google and the author of numerous publications on neural networks and related topics. Sutskever is a co-founder of DNNresearch and was named Canada’s first Google Fellow .... "  

Very good piece,  some very technical thoughts, but further Sutskever remarks:

" .... Unsupervised learning is always a means for some other end.   In supervised learning, the learning itself is what you care about. You’ve got your cost function, which you want to minimize. In unsupervised learning, the goal is always to help some other task, like classification or categorization. For example, I might ask a computer system to passively watch a lot of YouTube videos (so unsupervised learning happens here), then ask it to recognize objects with great accuracy (that’s the final supervised learning task).. .... " 

Follows with thoughts on the evolution and direction of deep learning. 

No comments: