ICLR 2021 accepted papers on Probabilistic and scientific modeling
88 of the 860 papers accepted at ICLR 2021 (10.2%) are on Probabilistic and scientific modeling, compared with 8.6% at ICLR 2020. They fall into 8 topics.
See ICLR 2021 in the interactive map
Submitting to ICLR? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Variational and generative sampling · 25 papers
- IDF++: Analyzing and Improving Integer Discrete Flows for Lossless Compression (35 citations)
- Neural Approximate Sufficient Statistics for Implicit Models (21 citations)
- not-MIWAE: Deep Generative Modelling with Missing not at Random Data (20 citations)
- Bidirectional Variational Inference for Non-Autoregressive Text-to-Speech (18 citations)
- Sliced Kernelized Stein Discrepancy (14 citations)
- Categorical Normalizing Flows via Continuous Transformations (13 citations)
- Disentangled Recurrent Wasserstein Autoencoder (13 citations)
- Learning a Latent Search Space for Routing Problems using Variational Autoencoders (12 citations)
- Property Controllable Variational Autoencoder via Invertible Mutual Dependence (9 citations)
- No MCMC for me: Amortized sampling for fast and stable training of energy-based models (7 citations)
- Offline Model-Based Optimization via Normalized Maximum Likelihood Estimation (7 citations)
- Rethinking the Role of Gradient-based Attribution Methods for Model Interpretability (7 citations)
and 13 more in the interactive map.
Molecules and proteins · 20 papers
- Transformer protein language models are unsupervised structure learners (372 citations)
- BERTology Meets Biology: Interpreting Attention in Protein Language Models (344 citations)
- Learning from Protein Structure with Geometric Vector Perceptrons (202 citations)
- MARS: Markov Molecular Sampling for Multi-objective Drug Discovery (29 citations)
- Learning Neural Generative Dynamics for Molecular Conformation Generation (28 citations)
- The role of Disentanglement in Generalisation (22 citations)
- Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures (18 citations)
- Predicting Infectiousness for Proactive Contact Tracing (8 citations)
- Learning Task-General Representations with Generative Neuro-Symbolic Modeling (5 citations)
- Trajectory Prediction using Equivariant Continuous Convolution (5 citations)
- Combining Physics and Machine Learning for Network Flow Estimation (4 citations)
- Neural Networks for Learning Counterfactual G-Invariances from Single Environments (3 citations)
and 8 more in the interactive map.
Dynamical systems · 12 papers
- Incorporating Symmetry into Deep Dynamics Models for Improved Generalization (78 citations)
- Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks (62 citations)
- Learning Mesh-Based Simulation with Graph Networks (49 citations)
- Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding (17 citations)
- Learning advanced mathematical computations from examples (7 citations)
- The geometry of integration in text classification RNNs (4 citations)
- Efficient Inference of Flexible Interaction in Spiking-neuron Networks (3 citations)
- Physics-aware, probabilistic model order reduction with guaranteed stability (2 citations)
- A teacher-student framework to distill future trajectories (1 citation)
- A Temporal Kernel Approach for Deep Learning with Continuous-time Information (1 citation)
- Identifying Physical Law of Hamiltonian Systems via Meta-Learning (1 citation)
- Nonseparable Symplectic Neural Networks
Uncertainty and Gaussian processes · 12 papers
- Getting a CLUE: A Method for Explaining Uncertainty Estimates (38 citations)
- Uncertainty in Gradient Boosting via Ensembles (38 citations)
- Uncertainty Estimation in Autoregressive Structured Prediction (31 citations)
- Contextual Dropout: An Efficient Sample-Dependent Dropout Module (11 citations)
- Bayesian Few-Shot Classification with One-vs-Each Pólya-Gamma Augmented Gaussian Processes (8 citations)
- A statistical theory of cold posteriors in deep neural networks (6 citations)
- Activation-level uncertainty in deep neural networks (3 citations)
- Exploring the Uncertainty Properties of Neural Networks’ Implicit Priors in the Infinite-Width Limit (3 citations)
- Large-width functional asymptotics for deep Gaussian neural networks (2 citations)
- Uncertainty Estimation and Calibration with Finite-State Probabilistic RNNs (2 citations)
- Tomographic Auto-Encoder: Unsupervised Bayesian Recovery of Corrupted Data
- Bayesian Context Aggregation for Neural Processes
Differential equations and PDEs · 9 papers
- Fourier Neural Operator for Parametric Partial Differential Equations (1,068 citations)
- Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting (113 citations)
- Learning Neural Event Functions for Ordinary Differential Equations (24 citations)
- PDE-Driven Spatiotemporal Disentanglement (11 citations)
- Learning continuous-time PDEs from sparse data with graph neural networks (8 citations)
- Learning Incompressible Fluid Dynamics from Scratch - Towards Fast, Differentiable Fluid Models that Generalize (6 citations)
- ResNet After All: Neural ODEs and Their Numerical Solution (6 citations)
- Go with the flow: Adaptive control for Neural ODEs (2 citations)
- Neural Delay Differential Equations
Causal discovery · 4 papers
- Interpretable Models for Granger Causality Using Self-explaining Neural Networks (28 citations)
- Debiasing Concept-based Explanations with Causal Analysis (8 citations)
- Latent Convergent Cross Mapping (5 citations)
- ANOCE: Analysis of Causal Effects with Multiple Mediators via Constrained Structural Learning (3 citations)
Bayesian optimization · 4 papers
- Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels (16 citations)
- Few-Shot Bayesian Optimization with Deep Kernel Surrogates (6 citations)
- AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly (5 citations)
- DynaTune: Dynamic Tensor Program Optimization in Deep Neural Network Compilation (4 citations)
