News for hackers. Stuff that matters.

nthngnss

Re:Learning to write programs that generate images on Wednesday March 28, 2018 @11:08PM (#16704231)
I would call "inverse graphics" a task. One can solve that task following different strategies. We demonstrate one way that uses RL and GANs and gives reasonably good results. For Omniglot, for example, there are works by Lake et al. that employ probabilistic perspective but the amount of hand-engineering involved makes their approach hard to apply to other tasks
Re:Learning to write programs that generate images on Tuesday March 27, 2018 @11:18PM (#16694966)
Yup, exactly! Btw, didn't realize those were your results (just mindlessly clicked on the link). Very cool stuff!
Re:Learning to write programs that generate images on Tuesday March 27, 2018 @10:50PM (#16694846)
While it is true that one can simulate painting by search (in some sense), the problem is that (naive) search doesn't always work (we have an example in the paper). Moreover, training an agent has a benefit of fast test-time inference (i.e., you give an image, and it paints it almost instantly). Of course, we haven't achieved the ultimate goal yet but it's a step in that direction.
Re:Learning to write programs that generate images on Tuesday March 27, 2018 @10:41PM (#16694801)
You'd be surprised to see that both "inverse graphics" and "analysis-by-synthesis" are mentioned in the paper
Re:Learning to write programs that generate images on Tuesday March 27, 2018 @05:11PM (#16692751)
The reason for that is that I chose a particular brush from the set of available brushes ('dry brush'). Since MNIST digits are quite sharp and opaque, the agent tries to achieve this by retracing the contours. I guess the remedy is to pick an appropriate brush style or make the agent choose it.