Search papers, labs, and topics across Lattice.
1
0
2
Forget full-batch gradients: this new pseudo-Langevin dynamics method lets you efficiently sample the parameter space of large neural networks using minibatches, scaling to millions of parameters.
