talks.md

Distributions

• Heavy tails in Bayesian NN [reporter: Marat Khusainov]
• Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks [reporter: "your name"]

Bayesian inference

• Hyperparameters: optimize, or integrate out? [reporter: "your name"]
• Learning Approximately Objective Priors [reporter: Galina Boeva]

Model complexity

• A widely applicable Bayesian information criterion [reporter: "your name"]
• The Description Length of Deep Learning Models [reporter: "Gavrilyuk Alexander"]

Variational inference 1

* Scalable marginal likelihood estimation for model selection in deep learning article [reporter: "your name"]

• Rényi divergence [reporter: Kseniia Petrushina]

Variationl inference 2

• Variational dropout [reporter: Eduard Vladimirov]
• Alpha-divergence [reporter: Boeva Galina]

Graphical models

• Learning to Discover Sparse Graphical Models [reporter: "your name"]
• Probabilistic circuits [reporter: "Parviz Karimov"]

Generative vs. Discriminative

• Hypertransformer: Model generation for supervised and semi-supervised few-shot learning [reporter: "Timofey Chernikov"]
• OUTRAGEOUSLY LARGE NEURAL NETWORKS: THE SPARSELY-GATED MIXTURE-OF-EXPERTS LAYER [reporter: Kseniia Petrushina]

Task 1 discussion (optional talk)

• Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks [reporter: "your name"]
• Hyperparameters: optimize, or integrate out? [reporter: "your name"]
• A widely applicable Bayesian information criterion [reporter: "your name"]
• Scalable marginal likelihood estimation for model selection in deep learning article [reporter: Maksim Tyurikov]

Generative models

• Dangers of Bayesian Model Averaging under Covariate Shift [reporter: "Timofey Chernikov"]
• Can You Trust Your Model’s Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift [reporter: "Dmitry Protasov"]

Hyperparameter optimization

• Weighted Random Search for Hyperparameter Optimization [reporter: Boeva Galina]
• c-TPE: Tree-Structured Parzen Estimator with Inequality Constraints for Expensive Hyperparameter Optimization [reporter: "your name"]

Gradient-based hyperparameter optimization

• Bayesian Optimization with Gradients [reporter: "your name"]
• Forward and Reverse Gradient-Based Hyperparameter Optimization [reporter: Maksim Tyurikov]

Task 2 discussion (optional talk)

• Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks [reporter: "your name"]
• Hyperparameters: optimize, or integrate out? [reporter: "your name"]
• A widely applicable Bayesian information criterion [reporter: "your name"]
• c-TPE: Tree-Structured Parzen Estimator with Inequality Constraints for Expensive Hyperparameter Optimization [reporter: "your name"]
• Bayesian Optimization with Gradients [reporter: "your name"]

Structure selection

• Neural Architecture Search without Training article [reporter: Dmitry Protasov]
• Bayesnas: A bayesian approach for neural architecture search article [reporter: "your name"]
• Bananas: Bayesian optimization with neural architectures for neural architecture search article [reporter: Boeva Galina]

Random processes and genetics for model generation

• AutoML-Zero: Evolving Machine Learning Algorithms From Scratch [reporter: Kseniia Petrushina]
• Proving the Lottery Ticket Hypothesis: Pruning is All You Need [reporter: "your name"]

Meta-optimization

• Generalized Inner Loop Meta-Learning [reporter: "your name"]
• HOW TO TRAIN YOUR MAML [reporter: Dmitry Protasov]

Multi-task learning

• Discovering Inductive Bias with Gibbs Priors: A Diagnostic Tool for Approximate Bayesian Inference [reporter: "your name"]
• Variational multi-task learning with gumbel-softmax priors [reporter: "your name"]

Knowledge transfer

• Dreaming to Distill: Data-free Knowledge Transfer via DeepInversion [reporter: Boeva Galina]
• Learning to select data for transfer learning with bayesian optimization [reporter: "your name"]
• Knowledge transfer via dense cross-layer mutual-distillation [reporter: "your name"]

Sampling

• Data augmentation in Bayesian neural networks and the cold posterior effect [reporter: "Marat Khusainov"]
• Evolutionary MCMC [reporter: "your name"]

Probabilistic metric spaces

• AN INDUCTIVE BIAS FOR DISTANCES: NEURAL NETS THAT RESPECT THE TRIANGLE INEQUALITY [reporter: "your name"]
• MsC: Siamese networks + prob. metric learning [reporter: "your name"]

Projection into latent space

• Neural operator search [reporter: "your name"]
• Super resolution neural operator [reporter: "your name"]

Model ensembles

• Functional MOE [reporter: "your name"]
• Neural ensemble search via Bayesian sampling [reporter: Kseniia Petrushina]

Gaussian processes

• The Variational Gaussian Process [reporter: Dmitry Protasov]
• State Space Expectation Propagation: Efficient Inference Schemes for Temporal Gaussian Processes [reporter: "your name"]