Next Generation Emulation banner

Machine-Learning (...) the Delusions of Big Data and Other Huge Engineering Efforts

420 views 2 replies 2 participants last post by  Milanó 
#1 ·
Pattern recognition and classification are closely tied to my field of research, so I found this interview a rather enjoyable read. If you're interested in machine learning (@GeominorAI), then I strongly recommend reading this.

The interviewee addresses multiple issues in the field, such as using terms which are no more than buzz words, focusing on the wrong approach (large data), among others.

http://spectrum.ieee.org/robotics/a...f-big-data-and-other-huge-engineering-efforts
 
#3 ·
I thought the article would focus more on the machine-learning problem, but I'm glad it went beyond that and I have to agree with his points.

I'm not a computer scientist, far from it, but I do know algorithm complexity and I have to agree with his view on the P = NP problem. If this statement is proven true, then it will have multiple important consequences, such as in the field of data security. However, by considering NP problem (e.g. the travelling salesman) such demonic existences, people think anything that's solvable in polynomial time is much better, which is true, but it won't matter much. Standard matrix multiplication is still O(n^3), which is infeasible for very large matrices such as those found in image processing and the best algorithm is about O(n^2.4). People in my field don't care that much whether an algorithm is exponential or polynomial, despite the former being much worst. All they want is something like O(n log n) like the FFT.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top