Designing models that are both expressive and preserve known invariances...
Training algorithms, broadly construed, are an essential part of every d...
Contrastive learning is a powerful framework for learning self-supervise...
Cutting planes are essential for solving mixed-integer linear problems
(...
Combinatorial optimization is a well-established area in operations rese...
Supervised learning can improve the design of state-of-the-art solvers f...
Skills or low-level policies in reinforcement learning are temporally
ex...
To perform counterfactual reasoning in Structural Causal Models (SCMs), ...
Training large-scale mixture of experts models efficiently on modern har...
Most data is automatically collected and only ever "seen" by algorithms....
Source code spends most of its time in a broken or incomplete state duri...
Latent variable models have been successfully applied in lossless compre...
We propose a general and scalable approximate sampling strategy for
prob...
Gradient estimation in models with discrete latent variables is a challe...
Propositional model counting or #SAT is the problem of computing the num...
The Gumbel-Max trick is the basis of many relaxed gradient estimators. T...
Selecting an optimizer is a central step in the contemporary deep learni...
Direct optimization is an appealing approach to differentiating through
...
The Variational Auto-Encoder (VAE) model has become widely popular as a ...
Deep latent variable models have become a popular model choice due to th...
We propose a family of optimization methods that achieve linear converge...
Deep neural networks excel at function approximation, yet they are typic...
We provide theoretical and empirical evidence that using tighter evidenc...
When used as a surrogate objective for maximum likelihood estimation in
...
Learning in models with discrete latent variables is challenging due to ...
The reparameterization trick enables optimizing large scale stochastic
c...
The game of Go is more challenging than other board games, due to the
di...
The problem of drawing samples from a discrete distribution can be conve...
We study the problem of building generative models of natural source cod...