Search papers, labs, and topics across Lattice.
2
0
2
Learning constant-depth circuits is possible even with correlated data from graphical models, challenging the reliance on i.i.d. assumptions in learning theory.
Sampling from diffusion models with quadratic rewards can be surprisingly hard: negative-definite tilts are intractable even in simple cases, while a new algorithm makes low-rank positive-definite tilts tractable.
