Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Thomas Hofmann

Thomas Hofmann

Analytic Insights into Structure and Rank of Neural Network Hessian Maps


Jul 01, 2021
Sidak Pal Singh, Gregor Bachmann, Thomas Hofmann

Figure 1 for Analytic Insights into Structure and Rank of Neural Network Hessian Maps
Figure 2 for Analytic Insights into Structure and Rank of Neural Network Hessian Maps
Figure 3 for Analytic Insights into Structure and Rank of Neural Network Hessian Maps
Figure 4 for Analytic Insights into Structure and Rank of Neural Network Hessian Maps

  Access Paper or Ask Questions

Precise characterization of the prior predictive distribution of deep ReLU networks


Jun 11, 2021
Lorenzo Noci, Gregor Bachmann, Kevin Roth, Sebastian Nowozin, Thomas Hofmann

Figure 1 for Precise characterization of the prior predictive distribution of deep ReLU networks
Figure 2 for Precise characterization of the prior predictive distribution of deep ReLU networks
Figure 3 for Precise characterization of the prior predictive distribution of deep ReLU networks

  Access Paper or Ask Questions

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


Jun 11, 2021
Lorenzo Noci, Kevin Roth, Gregor Bachmann, Sebastian Nowozin, Thomas Hofmann

Figure 1 for Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect
Figure 2 for Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect
Figure 3 for Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect
Figure 4 for Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect

  Access Paper or Ask Questions

Vanishing Curvature and the Power of Adaptive Methods in Randomly Initialized Deep Networks


Jun 07, 2021
Antonio Orvieto, Jonas Kohler, Dario Pavllo, Thomas Hofmann, Aurelien Lucchi

Figure 1 for Vanishing Curvature and the Power of Adaptive Methods in Randomly Initialized Deep Networks
Figure 2 for Vanishing Curvature and the Power of Adaptive Methods in Randomly Initialized Deep Networks
Figure 3 for Vanishing Curvature and the Power of Adaptive Methods in Randomly Initialized Deep Networks
Figure 4 for Vanishing Curvature and the Power of Adaptive Methods in Randomly Initialized Deep Networks

  Access Paper or Ask Questions

Uniform Convergence, Adversarial Spheres and a Simple Remedy


May 07, 2021
Gregor Bachmann, Seyed-Mohsen Moosavi-Dezfooli, Thomas Hofmann

Figure 1 for Uniform Convergence, Adversarial Spheres and a Simple Remedy
Figure 2 for Uniform Convergence, Adversarial Spheres and a Simple Remedy
Figure 3 for Uniform Convergence, Adversarial Spheres and a Simple Remedy
Figure 4 for Uniform Convergence, Adversarial Spheres and a Simple Remedy

  Access Paper or Ask Questions

Learning Generative Models of Textured 3D Meshes from Real-World Images


Mar 29, 2021
Dario Pavllo, Jonas Kohler, Thomas Hofmann, Aurelien Lucchi

Figure 1 for Learning Generative Models of Textured 3D Meshes from Real-World Images
Figure 2 for Learning Generative Models of Textured 3D Meshes from Real-World Images
Figure 3 for Learning Generative Models of Textured 3D Meshes from Real-World Images
Figure 4 for Learning Generative Models of Textured 3D Meshes from Real-World Images

  Access Paper or Ask Questions

Generative Minimization Networks: Training GANs Without Competition


Mar 23, 2021
Paulina Grnarova, Yannic Kilcher, Kfir Y. Levy, Aurelien Lucchi, Thomas Hofmann

Figure 1 for Generative Minimization Networks: Training GANs Without Competition
Figure 2 for Generative Minimization Networks: Training GANs Without Competition
Figure 3 for Generative Minimization Networks: Training GANs Without Competition
Figure 4 for Generative Minimization Networks: Training GANs Without Competition

  Access Paper or Ask Questions

SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German


Mar 21, 2021
Pelin Dogan-Schönberger, Julian Mäder, Thomas Hofmann

Figure 1 for SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German
Figure 2 for SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German
Figure 3 for SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German
Figure 4 for SwissDial: Parallel Multidialectal Corpus of Spoken Swiss German

  Access Paper or Ask Questions

Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization


Feb 23, 2021
Peiyuan Zhang, Antonio Orvieto, Hadi Daneshmand, Thomas Hofmann, Roy Smith

Figure 1 for Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization
Figure 2 for Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization
Figure 3 for Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization
Figure 4 for Revisiting the Role of Euler Numerical Integration on Acceleration and Stability in Convex Optimization
* 18 pages, 5 figures; Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS) 2021, San Diego, California, USA. PMLR: Volume 130 
  Access Paper or Ask Questions

Convolutional Generation of Textured 3D Meshes


Jun 13, 2020
Dario Pavllo, Graham Spinks, Thomas Hofmann, Marie-Francine Moens, Aurelien Lucchi

Figure 1 for Convolutional Generation of Textured 3D Meshes
Figure 2 for Convolutional Generation of Textured 3D Meshes
Figure 3 for Convolutional Generation of Textured 3D Meshes
Figure 4 for Convolutional Generation of Textured 3D Meshes

  Access Paper or Ask Questions

Theoretical Understanding of Batch-normalization: A Markov Chain Perspective


Mar 09, 2020
Hadi Daneshmand, Jonas Kohler, Francis Bach, Thomas Hofmann, Aurelien Lucchi

Figure 1 for Theoretical Understanding of Batch-normalization: A Markov Chain Perspective
Figure 2 for Theoretical Understanding of Batch-normalization: A Markov Chain Perspective
Figure 3 for Theoretical Understanding of Batch-normalization: A Markov Chain Perspective
Figure 4 for Theoretical Understanding of Batch-normalization: A Markov Chain Perspective

  Access Paper or Ask Questions

BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward


Mar 05, 2020
Florian Schmidt, Thomas Hofmann

Figure 1 for BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward
Figure 2 for BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward
Figure 3 for BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward
Figure 4 for BERT as a Teacher: Contextual Embeddings for Sequence-Level Reward

  Access Paper or Ask Questions

Controlling Style and Semantics in Weakly-Supervised Image Generation


Dec 06, 2019
Dario Pavllo, Aurelien Lucchi, Thomas Hofmann

Figure 1 for Controlling Style and Semantics in Weakly-Supervised Image Generation
Figure 2 for Controlling Style and Semantics in Weakly-Supervised Image Generation
Figure 3 for Controlling Style and Semantics in Weakly-Supervised Image Generation
Figure 4 for Controlling Style and Semantics in Weakly-Supervised Image Generation

  Access Paper or Ask Questions

Mixing of Stochastic Accelerated Gradient Descent


Oct 31, 2019
Peiyuan Zhang, Hadi Daneshmand, Thomas Hofmann

Figure 1 for Mixing of Stochastic Accelerated Gradient Descent
Figure 2 for Mixing of Stochastic Accelerated Gradient Descent

  Access Paper or Ask Questions

LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games


Sep 04, 2019
Leonard Adolphs, Thomas Hofmann

Figure 1 for LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games
Figure 2 for LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games
Figure 3 for LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games
Figure 4 for LeDeepChef: Deep Reinforcement Learning Agent for Families of Text-Based Games

  Access Paper or Ask Questions

Autoregressive Text Generation Beyond Feedback Loops


Aug 30, 2019
Florian Schmidt, Stephan Mandt, Thomas Hofmann

Figure 1 for Autoregressive Text Generation Beyond Feedback Loops
Figure 2 for Autoregressive Text Generation Beyond Feedback Loops
* emnlp camera ready 
  Access Paper or Ask Questions

Cosmological N-body simulations: a challenge for scalable generative models


Aug 15, 2019
Nathanaël Perraudin, Ankit Srivastava, Aurelien Lucchi, Tomasz Kacprzak, Thomas Hofmann, Alexandre Réfrégier

Figure 1 for Cosmological N-body simulations: a challenge for scalable generative models
Figure 2 for Cosmological N-body simulations: a challenge for scalable generative models
Figure 3 for Cosmological N-body simulations: a challenge for scalable generative models
Figure 4 for Cosmological N-body simulations: a challenge for scalable generative models

  Access Paper or Ask Questions

Adversarial Training Generalizes Data-dependent Spectral Norm Regularization


Jun 17, 2019
Kevin Roth, Yannic Kilcher, Thomas Hofmann

Figure 1 for Adversarial Training Generalizes Data-dependent Spectral Norm Regularization
Figure 2 for Adversarial Training Generalizes Data-dependent Spectral Norm Regularization
Figure 3 for Adversarial Training Generalizes Data-dependent Spectral Norm Regularization
Figure 4 for Adversarial Training Generalizes Data-dependent Spectral Norm Regularization

  Access Paper or Ask Questions

The Odds are Odd: A Statistical Test for Detecting Adversarial Examples


Feb 13, 2019
Kevin Roth, Yannic Kilcher, Thomas Hofmann

Figure 1 for The Odds are Odd: A Statistical Test for Detecting Adversarial Examples
Figure 2 for The Odds are Odd: A Statistical Test for Detecting Adversarial Examples
Figure 3 for The Odds are Odd: A Statistical Test for Detecting Adversarial Examples
Figure 4 for The Odds are Odd: A Statistical Test for Detecting Adversarial Examples

  Access Paper or Ask Questions

Evaluating GANs via Duality


Nov 13, 2018
Paulina Grnarova, Kfir Y Levy, Aurelien Lucchi, Nathanael Perraudin, Thomas Hofmann, Andreas Krause

Figure 1 for Evaluating GANs via Duality
Figure 2 for Evaluating GANs via Duality
Figure 3 for Evaluating GANs via Duality
Figure 4 for Evaluating GANs via Duality

  Access Paper or Ask Questions

Deep State Space Models for Unconditional Word Generation


Oct 28, 2018
Florian Schmidt, Thomas Hofmann

Figure 1 for Deep State Space Models for Unconditional Word Generation
Figure 2 for Deep State Space Models for Unconditional Word Generation
Figure 3 for Deep State Space Models for Unconditional Word Generation
Figure 4 for Deep State Space Models for Unconditional Word Generation
* NIPS camera-ready version 
  Access Paper or Ask Questions

Exponential convergence rates for Batch Normalization: The power of length-direction decoupling in non-convex optimization


Oct 06, 2018
Jonas Kohler, Hadi Daneshmand, Aurelien Lucchi, Ming Zhou, Klaus Neymeyr, Thomas Hofmann

Figure 1 for Exponential convergence rates for Batch Normalization: The power of length-direction decoupling in non-convex optimization
Figure 2 for Exponential convergence rates for Batch Normalization: The power of length-direction decoupling in non-convex optimization
Figure 3 for Exponential convergence rates for Batch Normalization: The power of length-direction decoupling in non-convex optimization
Figure 4 for Exponential convergence rates for Batch Normalization: The power of length-direction decoupling in non-convex optimization

  Access Paper or Ask Questions

Learning and Evaluating Sparse Interpretable Sentence Embeddings


Sep 25, 2018
Valentin Trifonov, Octavian-Eugen Ganea, Anna Potapenko, Thomas Hofmann

Figure 1 for Learning and Evaluating Sparse Interpretable Sentence Embeddings
Figure 2 for Learning and Evaluating Sparse Interpretable Sentence Embeddings
Figure 3 for Learning and Evaluating Sparse Interpretable Sentence Embeddings
Figure 4 for Learning and Evaluating Sparse Interpretable Sentence Embeddings
* Will be presented at the workshop "Analyzing and interpreting neural networks for NLP", collocated with the EMNLP 2018 conference in Brussels 
  Access Paper or Ask Questions

Fast Cosmic Web Simulations with Generative Adversarial Networks


Sep 20, 2018
Andres C. Rodriguez, Tomasz Kacprzak, Aurelien Lucchi, Adam Amara, Raphael Sgier, Janis Fluri, Thomas Hofmann, Alexandre Réfrégier

Figure 1 for Fast Cosmic Web Simulations with Generative Adversarial Networks
Figure 2 for Fast Cosmic Web Simulations with Generative Adversarial Networks
Figure 3 for Fast Cosmic Web Simulations with Generative Adversarial Networks
Figure 4 for Fast Cosmic Web Simulations with Generative Adversarial Networks

  Access Paper or Ask Questions

Escaping Saddles with Stochastic Gradients


Sep 16, 2018
Hadi Daneshmand, Jonas Kohler, Aurelien Lucchi, Thomas Hofmann

Figure 1 for Escaping Saddles with Stochastic Gradients
Figure 2 for Escaping Saddles with Stochastic Gradients
Figure 3 for Escaping Saddles with Stochastic Gradients
Figure 4 for Escaping Saddles with Stochastic Gradients

  Access Paper or Ask Questions

End-to-End Neural Entity Linking


Aug 29, 2018
Nikolaos Kolitsas, Octavian-Eugen Ganea, Thomas Hofmann

Figure 1 for End-to-End Neural Entity Linking
Figure 2 for End-to-End Neural Entity Linking
Figure 3 for End-to-End Neural Entity Linking
Figure 4 for End-to-End Neural Entity Linking
* Full paper at CoNLL 2018: Conference on Computational Natural Language Learning 
  Access Paper or Ask Questions

Hyperbolic Neural Networks


Jun 28, 2018
Octavian-Eugen Ganea, Gary Bécigneul, Thomas Hofmann

Figure 1 for Hyperbolic Neural Networks
Figure 2 for Hyperbolic Neural Networks
Figure 3 for Hyperbolic Neural Networks

  Access Paper or Ask Questions

Local Saddle Point Optimization: A Curvature Exploitation Approach


Jun 26, 2018
Leonard Adolphs, Hadi Daneshmand, Aurelien Lucchi, Thomas Hofmann

Figure 1 for Local Saddle Point Optimization: A Curvature Exploitation Approach
Figure 2 for Local Saddle Point Optimization: A Curvature Exploitation Approach
Figure 3 for Local Saddle Point Optimization: A Curvature Exploitation Approach
Figure 4 for Local Saddle Point Optimization: A Curvature Exploitation Approach

  Access Paper or Ask Questions