= Gradient Estimation Using Stochastic Computation Graphs by John Schulman and others
{c}
{tag=Work by John Schulman}
{title2=2015}
{title2=arXiv:1506.05254}
https://proceedings.neurips.cc/paper/2015/file/de03beffeed9da5f3639a621bcab5dd4-Paper.pdf[Primary conference paper]. Section 3.1 and Appendix A derive the general <gradient> identity; Section 4 treats baselines. The construction combines random-choice <probabilities> and differentiable deterministic paths in a <stochastic computation graph>.
Back to article page
