Edinburgh Research Explorer Continuous Relaxations for Discrete Hamiltonian Monte Carlo

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Generalizing Hamiltonian Monte Carlo with Neural Networks (2017, methodology): “…dynamics (Tripuraneni et al., 2016); with variants of AIS adapted to HMC proposals (Sohl-Dickstein & Culpepper, 2012); with the extension of HMC to discrete state spaces (Zhang et al., 2012); and with the use of alternative Hamiltonian integrators (Creutz & Gocksch, 1989; Chao et al., 2015).”
DVAE#: Discrete Variational Autoencoders with Relaxed Boltzmann Priors (2018, methodology): “Instead of applying mean ﬁeld, we remove these interactions using the Gaussian integral trick [41].”
Neural Network Renormalization Group (2018, methodology): “By diagonalizing (K + αI) one obtains a model where each continuous variable couples to one Fourier component of the original Ising spins [63, 64].”
Monge-Ampère Flow for Generative Modeling (2018, background): “listic model. To predict the thermodynamics accurately, the ﬂow has to capture long-range correlations and rich physics due to the critical ﬂuctuation. In the continuous representation Fisher (1983); Zhang et al. (2012); Li & Wang (2018), the energy function of the Ising model reads (Appendix B)”