Search Results for author: Sebastian Nowozin

Found 54 papers, 24 papers with code

Memory Efficient Meta-Learning with Large Images

2 code implementations2 Jul 2021 John Bronskill, Daniela Massiceti, Massimiliano Patacchiola, Katja Hofmann, Sebastian Nowozin, Richard E. Turner

This limitation arises because a task's entire support set, which can contain up to 1000 images, must be processed before an optimization step can be taken.

Meta-Learning Transfer Learning

Precise characterization of the prior predictive distribution of deep ReLU networks

no code implementations11 Jun 2021 Lorenzo Noci, Gregor Bachmann, Kevin Roth, Sebastian Nowozin, Thomas Hofmann

Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture.

Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect

no code implementations11 Jun 2021 Lorenzo Noci, Kevin Roth, Gregor Bachmann, Sebastian Nowozin, Thomas Hofmann

The dataset curation hypothesis of Aitchison (2020): we show empirically that the CPE does not arise in a real curated data set but can be produced in a controlled experiment with varying curation strength.

Data Augmentation

TaskNorm: Rethinking Batch Normalization for Meta-Learning

2 code implementations ICML 2020 John Bronskill, Jonathan Gordon, James Requeima, Sebastian Nowozin, Richard E. Turner

Modern meta-learning approaches for image classification rely on increasingly deep networks to achieve state-of-the-art performance, making batch normalization an essential component of meta-learning pipelines.

Classification General Classification +2

How Good is the Bayes Posterior in Deep Neural Networks Really?

1 code implementation ICML 2020 Florian Wenzel, Kevin Roth, Bastiaan S. Veeling, Jakub Świątkowski, Linh Tran, Stephan Mandt, Jasper Snoek, Tim Salimans, Rodolphe Jenatton, Sebastian Nowozin

In this work we cast doubt on the current understanding of Bayes posteriors in popular deep neural networks: we demonstrate through careful MCMC sampling that the posterior predictive induced by the Bayes posterior yields systematically worse predictions compared to simpler methods including point estimates obtained from SGD.

Bayesian Inference

Icebreaker: Element-wise Efficient Information Acquisition with a Bayesian Deep Latent Gaussian Model

1 code implementation NeurIPS 2019 Wenbo Gong, Sebastian Tschiatschek, Sebastian Nowozin, Richard E. Turner, José Miguel Hernández-Lobato, Cheng Zhang

In this paper, we address the ice-start problem, i. e., the challenge of deploying machine learning models when only a little or no training data is initially available, and acquiring each feature element of data is associated with costs.

Imputation Recommendation Systems

Independent Subspace Analysis for Unsupervised Learning of Disentangled Representations

no code implementations5 Sep 2019 Jan Stühmer, Richard E. Turner, Sebastian Nowozin

Second, we demonstrate that the proposed prior encourages a disentangled latent representation which facilitates learning of disentangled representations.

Variational Inference

Icebreaker: Element-wise Active Information Acquisition with Bayesian Deep Latent Gaussian Model

1 code implementation13 Aug 2019 Wenbo Gong, Sebastian Tschiatschek, Richard Turner, Sebastian Nowozin, José Miguel Hernández-Lobato, Cheng Zhang

In this paper we introduce the ice-start problem, i. e., the challenge of deploying machine learning models when only little or no training data is initially available, and acquiring each feature element of data is associated with costs.

Active Learning Imputation +1

Fast and Flexible Multi-Task Classification Using Conditional Neural Adaptive Processes

2 code implementations NeurIPS 2019 James Requeima, Jonathan Gordon, John Bronskill, Sebastian Nowozin, Richard E. Turner

We introduce a conditional neural process based approach to the multi-task classification setting for this purpose, and establish connections to the meta-learning and few-shot learning literature.

Active Learning Continual Learning +3

Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift

2 code implementations NeurIPS 2019 Yaniv Ovadia, Emily Fertig, Jie Ren, Zachary Nado, D. Sculley, Sebastian Nowozin, Joshua V. Dillon, Balaji Lakshminarayanan, Jasper Snoek

Modern machine learning methods including deep learning have achieved great success in predictive accuracy for supervised learning tasks, but may still fall short in giving useful estimates of their predictive {\em uncertainty}.

Probabilistic Deep Learning

ISA-VAE: Independent Subspace Analysis with Variational Autoencoders

no code implementations ICLR 2019 Jan Stühmer, Richard Turner, Sebastian Nowozin

Extensive quantitative and qualitative experiments demonstrate that the proposed prior mitigates the trade-off introduced by modified cost functions like beta-VAE and TCVAE between reconstruction loss and disentanglement.

Variational Inference

Contextual Face Recognition with a Nested-Hierarchical Nonparametric Identity Model

no code implementations19 Nov 2018 Daniel C. Castro, Sebastian Nowozin

Current face recognition systems typically operate via classification into known identities obtained from supervised identity annotations.

Face Recognition General Classification

Deterministic Variational Inference for Robust Bayesian Neural Networks

2 code implementations ICLR 2019 Anqi Wu, Sebastian Nowozin, Edward Meeds, Richard E. Turner, José Miguel Hernández-Lobato, Alexander L. Gaunt

We provide two innovations that aim to turn VB into a robust inference tool for Bayesian neural networks: first, we introduce a novel deterministic method to approximate moments in neural networks, eliminating gradient variance; second, we introduce a hierarchical prior for parameters and a novel Empirical Bayes procedure for automatically selecting prior variances.

Variational Inference

EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE

1 code implementation ICLR 2019 Chao Ma, Sebastian Tschiatschek, Konstantina Palla, José Miguel Hernández-Lobato, Sebastian Nowozin, Cheng Zhang

Many real-life decision-making situations allow further relevant information to be acquired at a specific cost, for example, in assessing the health status of a patient we may decide to take additional measurements such as diagnostic tests or imaging scans before making a final assessment.

Decision Making Experimental Design

From Face Recognition to Models of Identity: A Bayesian Approach to Learning about Unknown Identities from Unsupervised Data

no code implementations ECCV 2018 Daniel C. Castro, Sebastian Nowozin

There are two problems with this current paradigm: (1) current systems are unable to benefit from unlabelled data which may be available in large quantities; and (2) current systems equate successful recognition with labelling a given input image.

Face Recognition General Classification

Meta-Learning Probabilistic Inference For Prediction

1 code implementation ICLR 2019 Jonathan Gordon, John Bronskill, Matthias Bauer, Sebastian Nowozin, Richard E. Turner

2) We introduce VERSA, an instance of the framework employing a flexible and versatile amortization network that takes few-shot learning datasets as inputs, with arbitrary numbers of shots, and outputs a distribution over task-specific parameters in a single forward pass.

Few-Shot Learning

Adversarially Robust Training through Structured Gradient Regularization

no code implementations22 May 2018 Kevin Roth, Aurelien Lucchi, Sebastian Nowozin, Thomas Hofmann

We propose a novel data-dependent structured gradient regularizer to increase the robustness of neural networks vis-a-vis adversarial perturbations.

Deep Directional Statistics: Pose Estimation with Uncertainty Quantification

1 code implementation ECCV 2018 Sergey Prokudin, Peter Gehler, Sebastian Nowozin

However, in challenging imaging conditions such as on low-resolution images or when the image is corrupted by imaging artifacts, current systems degrade considerably in accuracy.

Pose Estimation Probabilistic Deep Learning

Which Training Methods for GANs do actually Converge?

9 code implementations ICML 2018 Lars Mescheder, Andreas Geiger, Sebastian Nowozin

In this paper, we show that the requirement of absolute continuity is necessary: we describe a simple yet prototypical counterexample showing that in the more realistic case of distributions that are not absolutely continuous, unregularized GAN training is not always convergent.

Debiasing Evidence Approximations: On Importance-weighted Autoencoders and Jackknife Variational Inference

1 code implementation ICLR 2018 Sebastian Nowozin

The importance-weighted autoencoder (IWAE) approach of Burda et al. defines a sequence of increasingly tighter bounds on the marginal likelihood of latent variable models.

Latent Variable Models Variational Inference

Generative Entity Networks: Disentangling Entitites and Attributes in Visual Scenes using Partial Natural Language Descriptions

no code implementations ICLR 2018 Charlie Nash, Sebastian Nowozin, Nate Kushman

Using the Shapeworld dataset, we show that our representation both enables a better generative model of images, leading to higher quality image samples, as well as creating more semantically useful representations that improve performance over purely dicriminative models on a simple natural language yes/no question answering task.

Question Answering

Hybrid VAE: Improving Deep Generative Models using Partial Observations

no code implementations30 Nov 2017 Sergey Tulyakov, Andrew Fitzgibbon, Sebastian Nowozin

We show that such a combination is beneficial because the unlabeled data acts as a data-driven form of regularization, allowing generative models trained on few labeled samples to reach the performance of fully-supervised generative models trained on much larger datasets.

PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples

1 code implementation ICLR 2018 Yang Song, Taesup Kim, Sebastian Nowozin, Stefano Ermon, Nate Kushman

Adversarial perturbations of normal images are usually imperceptible to humans, but they can seriously confuse state-of-the-art machine learning models.

Two-sample testing

Dynamic Time-Of-Flight

no code implementations CVPR 2017 Michael Schober, Amit Adam, Omer Yair, Shai Mazor, Sebastian Nowozin

Operating in this mode the camera essentially forgets all information previously captured - and performs depth inference from scratch for every frame.

The Atari Grand Challenge Dataset

2 code implementations31 May 2017 Vitaly Kurin, Sebastian Nowozin, Katja Hofmann, Lucas Beyer, Bastian Leibe

Recent progress in Reinforcement Learning (RL), fueled by its combination, with Deep Learning has enabled impressive results in learning to interact with complex virtual environments, yet real-world applications of RL are still scarce.

Imitation Learning

The Numerics of GANs

4 code implementations NeurIPS 2017 Lars Mescheder, Sebastian Nowozin, Andreas Geiger

In this paper, we analyze the numerics of common algorithms for training Generative Adversarial Networks (GANs).

Stabilizing Training of Generative Adversarial Networks through Regularization

1 code implementation NeurIPS 2017 Kevin Roth, Aurelien Lucchi, Sebastian Nowozin, Thomas Hofmann

Deep generative models based on Generative Adversarial Networks (GANs) have demonstrated impressive sample quality but in order to work they require a careful choice of architecture, parameter initialization, and selection of hyper-parameters.

Image Generation

Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observations

2 code implementations24 May 2017 Diane Bouchacourt, Ryota Tomioka, Sebastian Nowozin

We would like to learn a representation of the data which decomposes an observation into factors of variation which we can independently control.

Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks

1 code implementation ICML 2017 Lars Mescheder, Sebastian Nowozin, Andreas Geiger

We show that in the nonparametric limit our method yields an exact maximum-likelihood assignment for the parameters of the generative model, as well as the exact posterior distribution over the latent variables given an observation.

Latent Variable Models

DISCO Nets : DISsimilarity COefficients Networks

no code implementations NeurIPS 2016 Diane Bouchacourt, Pawan K. Mudigonda, Sebastian Nowozin

We present a new type of probabilistic model which we call DISsimilarity COefficient Networks (DISCO Nets).

Memory Lens: How Much Memory Does an Agent Use?

no code implementations21 Nov 2016 Christoph Dann, Katja Hofmann, Sebastian Nowozin

The study of memory as information that flows from the past to the current action opens avenues to understand and improve successful reinforcement learning algorithms.

Probabilistic Duality for Parallel Gibbs Sampling without Graph Coloring

no code implementations21 Nov 2016 Lars Mescheder, Sebastian Nowozin, Andreas Geiger

We present a new notion of probabilistic duality for random variables involving mixture distributions.

DSAC - Differentiable RANSAC for Camera Localization

4 code implementations CVPR 2017 Eric Brachmann, Alexander Krull, Sebastian Nowozin, Jamie Shotton, Frank Michel, Stefan Gumhold, Carsten Rother

The most promising approach is inspired by reinforcement learning, namely to replace the deterministic hypothesis selection by a probabilistic selection for which we can derive the expected loss w. r. t.

Camera Localization Visual Localization

DeepCoder: Learning to Write Programs

3 code implementations7 Nov 2016 Matej Balog, Alexander L. Gaunt, Marc Brockschmidt, Sebastian Nowozin, Daniel Tarlow

We develop a first line of attack for solving programming competition-style problems from input-output examples using deep learning.

Enumerative Search

DISCO Nets: DISsimilarity COefficient Networks

no code implementations8 Jun 2016 Diane Bouchacourt, M. Pawan Kumar, Sebastian Nowozin

We present a new type of probabilistic model which we call DISsimilarity COefficient Networks (DISCO Nets).

f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization

1 code implementation NeurIPS 2016 Sebastian Nowozin, Botond Cseke, Ryota Tomioka

Generative neural samplers are probabilistic models that implement sampling using feedforward neural networks: they take a random input vector and produce a sample from a probability distribution defined by the network weights.

Model-Based Tracking at 300Hz Using Raw Time-of-Flight Observations

no code implementations ICCV 2015 Jan Stuhmer, Sebastian Nowozin, Andrew Fitzgibbon, Richard Szeliski, Travis Perry, Sunil Acharya, Daniel Cremers, Jamie Shotton

In this paper, we show how to perform model-based object tracking which allows to reconstruct the object's depth at an order of magnitude higher frame-rate through simple modifications to an off-the-shelf depth camera.

Object Tracking

Entropy-Based Latent Structured Output Prediction

no code implementations ICCV 2015 Diane Bouchacourt, Sebastian Nowozin, M. Pawan Kumar

To this end, we propose a novel prediction criterion that includes as special cases all previous prediction criteria that have been used in the literature.

Structured Prediction

Bayesian Time-of-Flight for Realtime Shape, Illumination and Albedo

no code implementations22 Jul 2015 Amit Adam, Christoph Dann, Omer Yair, Shai Mazor, Sebastian Nowozin

We propose a computational model for shape, illumination and albedo inference in a pulsed time-of-flight (TOF) camera.

Optimal Decisions from Probabilistic Models: The Intersection-over-Union Case

no code implementations CVPR 2014 Sebastian Nowozin

A probabilistic model allows us to reason about the world and make statistically optimal decisions using Bayesian decision theory.

Decision Making Semantic Segmentation

Efficient Nonlinear Markov Models for Human Motion

no code implementations CVPR 2014 Andreas M. Lehrmann, Peter V. Gehler, Sebastian Nowozin

The use of hidden variables makes them expressive models, but inference is only approximate and requires procedures such as particle filters or Markov chain Monte Carlo methods.

Action Recognition Motion Capture

Cascades of Regression Tree Fields for Image Restoration

no code implementations8 Apr 2014 Uwe Schmidt, Jeremy Jancsary, Sebastian Nowozin, Stefan Roth, Carsten Rother

We posit two reasons for this: First, the blur kernel is often only known at test time, requiring any discriminative approach to cope with considerable variability.

Blind Image Deblurring Image Denoising +1

A Comparative Study of Modern Inference Techniques for Structured Discrete Energy Minimization Problems

no code implementations2 Apr 2014 Jörg H. Kappes, Bjoern Andres, Fred A. Hamprecht, Christoph Schnörr, Sebastian Nowozin, Dhruv Batra, Sungwoong Kim, Bernhard X. Kausler, Thorben Kröger, Jan Lellmann, Nikos Komodakis, Bogdan Savchynskyy, Carsten Rother

However, on new and challenging types of models our findings disagree and suggest that polyhedral methods and integer programming solvers are competitive in terms of runtime and solution quality over a large range of model types.

Bayesian Inference for NMR Spectroscopy with Applications to Chemical Quantification

no code implementations14 Feb 2014 Andrew Gordon Wilson, Yuting Wu, Daniel J. Holland, Sebastian Nowozin, Mick D. Mantle, Lynn F. Gladden, Andrew Blake

Nuclear magnetic resonance (NMR) spectroscopy exploits the magnetic properties of atomic nuclei to discover the structure, reaction state and chemical environment of molecules.

Bayesian Inference

The Informed Sampler: A Discriminative Approach to Bayesian Inference in Generative Computer Vision Models

1 code implementation4 Feb 2014 Varun Jampani, Sebastian Nowozin, Matthew Loper, Peter V. Gehler

Computer vision is hard because of a large variability in lighting, shape, and texture; in addition the image signal is non-additive due to occlusion.

Bayesian Inference

Decision Jungles: Compact and Rich Models for Classification

no code implementations NeurIPS 2013 Jamie Shotton, Toby Sharp, Pushmeet Kohli, Sebastian Nowozin, John Winn, Antonio Criminisi

Randomized decision trees and forests have a rich history in machine learning and have seen considerable success in application, perhaps particularly so for computer vision.

Classification General Classification

Improved Information Gain Estimates for Decision Tree Induction

no code implementations18 Jun 2012 Sebastian Nowozin

Ensembles of classification and regression trees remain popular machine learning methods because they define flexible non-parametric models that predict well and are computationally efficient both during training and testing.

Classification General Classification

Cannot find the paper you are looking for? You can Submit a new open access paper.