Sup guys. ML noob here.
Is there such a thing as the opposite of pattern recognition? is there a proper name for it?
E.g., in hand-written digit recognition (supervised or unsupervised) there is the pattern '1', and there is the idea '1' behind the pattern. And the purpose of ML code is to recognize the idea from the pattern.
But I want a code, that has learnt the pattern, to generate patterns for that idea, such that no two generated patterns are alike.
Any ideas?
[link][4 comments]