Normalizing flows are explicit likelihood models using invertible neural networks to construct flexible probability distributions of high-dimensional data. Compared to other generative models, the main advantage of normalizing flows is that they can offer exact and efficient likelihood computation and data generation. Since their recent introduction, flow-based models have seen a significant resurgence of interest in the machine learning community. As a result, powerful flow-based models have been developed, with successes in density estimation, variational inference, and generative modeling of images, audio and video.
This workshop is the 2nd iteration of the ICML 2019 workshop on Invertible Neural Networks and Normalizing Flows. While the main goal of last yearâ€™s workshop was to make flow-based models more accessible to the general machine learning community, as the field is moving forward, we believe there is now a need to consolidate recent progress and connect ideas from related fields. In light of the interpretation of latent variable models and autoregressive models as flows, this year we expand the scope of the workshop and consider likelihood-based models more broadly, including flow-based models, latent variable models and autoregressive models. We encourage the researchers to use these models in conjunction to exploit the their benefits at once, and to work together to resolve some common issues of likelihood-based methods, such as mis-calibration of out-of-distribution uncertainty.
Opening remarks (Introduction) | |
Invited talk 1: Unifying VAEs and Flows (talk) | |
Q&A with Max Welling (Q&A) | |
Spotlight talk: Neural Manifold Ordinary Differential Equations (talk) | |
Spotlight talk: The Convolution Exponential (talk) | |
Spotlight talk: WaveNODE: A Continuous Normalizing Flow for Speech Synthesis (talk) | |
Spotlight talk: Neural Ordinary Differential Equations on Manifolds (talk) | |
Poster session 1 (poster) | |
Invited talk 2: Detecting Distribution Shift with Deep Generative Models (talk) | |
Q&A with Eric Nalisnick (Q&A) | |
Invited talk 3: Representational limitations of invertible models (talk) | |
Q&A with Emilien Dupont (Q&A) | |
Spotlight talk: You say Normalizing Flows I see Bayesian Networks (talk) | |
Spotlight talk: Variational Inference with Continuously-Indexed Normalizing Flows (talk) | |
Spotlight talk: NOTAGAN: Flows for the data manifold (talk) | |
Spotlight talk: Ordering Dimensions with Nested Dropout Normalizing Flows (talk) | |
Spotlight talk: The Lipschitz Constant of Self-Attention (talk) | |
Spotlight talk: Autoregressive flow-based causal discovery and inference (talk) | |
Lunch break (break) | |
Invited talk 4: Divergence Measures in Variational Inference and How to Choose Them (talk) | |
Q&A with Cheng Zhang (Q&A) | |
Invited talk 5: Adversarial Learning of Prescribed Generative Models (talk) | |
Q&A with Adji Bousso Dieng (Q&A) | |
Contributed talk: Modeling Continuous Stochastic Processes with Dynamic Normalizing Flows (talk) | |
Q&A with authors of contributed talk (Q&A) | |
Invited talk 6: Likelihood Models for Science (talk) | |
Q&A with Kyle Cranmer (Q&A) | |
Invited talk 7: Flows in Probabilistic Modeling & Inference (talk) | |
Q&A with Martin Jankowiak (Q&A) | |
Contributed talk: Learning normalizing flows from Entropy-Kantorovich potentials (talk) | |
Q&A with authors of contributed talk (Q&A) | |
Poster session 2 (poster) | |
Poster presentation: Scaling RBMs to High Dimensional Data with Invertible Neural Networks (talk) | |
Poster presentation: On the Variational Posterior of Dirichlet Process Deep Latent Gaussian Mixture Models (talk) | |
Poster presentation: A Fourier State Space Model for Bayesian ODE Filters (talk) | |
Poster presentation: Woodbury Transformations for Deep Generative Flows (talk) | |
Poster presentation: Metropolized Flow: from Invertible Flow to MCMC (talk) | |
Poster presentation: Why Normalizing Flows Fail to Detect Out-of-Distribution Data (talk) | |
Poster presentation: MoFlow: An Invertible Flow Model for Molecular Graph Generation (talk) | |
Link: Poster presentations and zoom links (Link) | |
Poster presentation: Stochastic Normalizing Flows (talk) | |
Poster presentation: TraDE: Transformers for Density Estimation (talk) | |
Poster presentation: WeakFlow: Iterative Invertible Distribution Transformations via Weak Destructive Flows (talk) | |
Poster presentation: Flow-based SVDD for anomaly detection (talk) | |
Poster presentation: Sequential Autoregressive Flow-Based Policies (talk) | |
Poster presentation: Black-box Adversarial Example Generation with Normalizing Flows (talk) | |
Poster presentation: Relative gradient optimization of the Jacobian term in unsupervised deep learning (talk) | |
Poster presentation: Deep Generative Video Compression with Temporal Autoregressive Transforms (talk) | |
Poster presentation: Normalizing Flows Across Dimensions (talk) | |
Poster presentation: Model-Agnostic Searches for New Physics with Normalizing Flows (talk) | |
Poster presentation: Differentially Private Normalizing Flows for Privacy-Preserving Density Estimation (talk) | |
Poster presentation: Density Deconvolution with Normalizing Flows (talk) | |
Link: Slack (link) | |
Poster presentation: Conditional Normalizing Flows for Low-Dose Computed Tomography Image Reconstruction (talk) | |
Poster presentation: Consistency Regularization for Variational Auto-encoders (talk) | |
Poster presentation: Super-resolution Variational Auto-Encoders (talk) | |
Poster presentation: Normalizing Flows with Multi-Scale Autoregressive Priors (talk) | |
Poster presentation: The Power Spherical distribution (talk) | |
Poster presentation: Robust model training and generalisation with Studentising flows (talk) | |
Poster presentation: Faster Orthogonal Parameterization with Householder Matrices (talk) | |
Poster presentation: Time Series Decomposition with Slow Flows (talk) | |
Poster presentation: Quasi-Autoregressive Residual (QuAR) Flows (talk) | |
Poster presentation: Exhaustive Neural Importance Sampling applied to Monte Carlo event generation (talk) | |
Poster presentation: Improving Sample Quality by Training and Sampling from Latent Energy (talk) | |