4 ms·
Paper: http://www.mit.edu/~ilkery/papers/phys_nips.pdf http://www.mit.edu/~ilkery/papers/phys_nips.pdf
by nharada 10y ago
Paper: http://www.mit.edu/~ilkery/papers/phys_nips.pdf http://www.mit.edu/~ilkery/papers/phys_nips.pdf
- eli_gottlieb 10y agoAka: "Another probabilistic programming paper." They hacked on a 3D physics engine to make it capable of Monte Carlo sampling and inference, and then stuck a deep neural net on the front to do some parameter inference more easily.