This is an old revision of the document!
Applications
GANs / unsupervised
IntroVAE: Introspective Variational Autoencoders for Photographic Image Synthesis Wasserstein GAN Text adaptive GAN: Manipulating images with natural language
Network design
Attention is all you need Neural Ordinary Differential Equations - https://arxiv.org/pdf/1806.07366.pdf Reversible neural networks - https://arxiv.org/abs/1807.03039 - https://arxiv.org/abs/1605.08803
Foundations / Philosophy
Troubling trends in ML scholarship - https://arxiv.org/pdf/1807.03341 A Theory of Local Learning, the Learning Channel, and the Optimality of Backpropagation - https://arxiv.org/pdf/1506.06472 Why and When Can Deep -- but Not Shallow -- Networks Avoid the Curse of Dimensionality: a Review - https://arxiv.org/pdf/1611.00740
RL
Curiosity-driven exploration by self-supervised prediction Diversity is all you need: Learning skills without a reward function - http://arxiv.org/pdf/1802.06070v6.pdf World Models - https://arxiv.org/pdf/1803.10122v4.pdf
Graph networks
Graph Neural Networks: A Review of Methods and Applications - https://arxiv.org/abs/1812.08434 Relational inductive biases, deep learning, and graph networks - https://arxiv.org/pdf/1806.01261
Optimization / training
Averaging weights leads to wider optima and better generalization - http://arxiv.org/pdf/1803.05407v2.pdf The loss surface of multilayer networks - https://arxiv.org/pdf/1412.0233 Visualizing The Loss Landscape of Neural Nets - https://arxiv.org/pdf/1712.09913v3.pdf The Matrix Calculus You Need For Deep Learning - https://arxiv.org/pdf/1802.01528v3.pdf Group Norm - https://arxiv.org/pdf/1803.08494v3.pdf Kalman Normalization: Normalizing internal representations across network layers MetaReg: towards Domain Generalization using meta-regularization AutoAugment - https://arxiv.org/abs/1805.09501 A Disciplined Approach To Neural Network Hyper-Parameters: part 1 - http://arxiv.org/pdf/1803.09820v2.pdf (Direct) Feedback alignment
Geometric deep learning
Geometric deep learning: going beyond Euclidean data- https://arxiv.org/pdf/1611.08097.pdf Convolutional Neural Networks on Surfaces via Seamless Toric Covers SchNet: A continuous-filter convolutional neural network for modeling quantum interactions Deriving Neural Architectures from Sequence and Graph Kernels CayleyNets: Graph convolutional neural networks with complex rational spectral filters Deep Functional Maps: Structured Prediction for Dense Shape Correspondence Geometric matrix completion with recurrent multi-graph neural networks Neural Message Passing for Quantum Chemistry Deep Learning on Lie Groups for Skeleton-based Action Recognition
Other
Bayesian neural networks?