In general, TensorFlow and other automatic differentiation frameworks (Chainer, Torch, Theano, etc.) operate on gradient descent of smooth functions. If you can recast your combinatorial problem into a smooth function, then these frameworks may be applicable. However, most combinatorial optimization problems aren't readily transformed into smooth functions, so this is unlikely to be straightforward.
You can use the tensor formalism to express combinatorial problems, but these will not be smooth, so gradient descent algorithms will not work. However, you can turn such boolean tensor problems into huge SAT problems and solve them with SAT solvers. See for example here: https://github.com/mmaroti/uasat
2 comments
[ 6.0 ms ] story [ 30.9 ms ] thread