Search papers, labs, and topics across Lattice.
University of Texas at Austin
5
0
4
Polynomial-time sampling in spin glasses and sparse Bayesian regression is tractable deep into low-temperature regimes, slashing the measurement barrier for spike-and-slab posteriors from $k^3$ to $k^{3/2}$ while approaching the Almeida鈥揟houless phase boundary.
Achieving optimal recalibration rates in online prediction could redefine performance benchmarks in adaptive learning systems.
Weakening error assumptions reveals that unbiased sampling may be impossible, reshaping our understanding of tractable sampling methods.
Forget single-objective optimization鈥攖his work cracks omniprediction in multiclass settings, opening the door to algorithms that are robust across diverse loss functions and comparator classes.
Sparse recovery requires quadratically more samples in adaptive settings than oblivious ones, a surprising divergence from the well-understood $\ell_2$ norm.
