1
$\begingroup$

I was interested in learning more about program induction and program synthesis. I was especially interested in learning about how to form program from data for example. My background is in traditional ML+Neural Nets so I have no idea were to start...I am familiar with Program Sythesis by Sketching but beyond that don't know what would be a good say, introduction textbook to the field. Any pointers to get started would be good! Especially if they are the standard foundations for program synthesis/induction.

Of course I saw there are wikipedia articles but they didn't seem like a great reference to learn the foundations and then move forward...

$\endgroup$
4
  • $\begingroup$ One approach taken at Microsoft research is through deductive synthesis: microsoft.github.io/prose $\endgroup$
    – 2bigpigs
    Commented Apr 26, 2018 at 6:24
  • $\begingroup$ Another one is based more on Inductive Logic Programming. I haven't read this paper yet but since it's written by Stephen Muggleton, I assume it'll be good: pdfs.semanticscholar.org/6a66/… $\endgroup$
    – 2bigpigs
    Commented Apr 26, 2018 at 6:27
  • $\begingroup$ I can't seem to edit my earlier comment. Since you know Neural nets, Here's a paper which uses RNN to guide the search ( program tree generation ) in Prose: openreview.net/forum?id=rywDjg-RW This does compare the approach to a couple of statistical methods as well. ( I personally share the sentiment of the "Although the search method chosen was reasonable, the only real innovation here is to use the LSTM to learn a search heuristic." review on this one, but it's good food for thought. can probably used in the other direction to use a grammar to constrain an NN $\endgroup$
    – 2bigpigs
    Commented Apr 26, 2018 at 6:31
  • $\begingroup$ Better link to MSR's PROSE: microsoft.com/en-us/research/publication/… $\endgroup$
    – 2bigpigs
    Commented Apr 26, 2018 at 6:35

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.