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

Add to Chrome

Add to Firefox

Picture for Kenji Kawaguchi

Kenji Kawaguchi

Meta-learning PINN loss functions

Jul 12, 2021
Apostolos F Psaros, Kenji Kawaguchi, George Em Karniadakis

Figure 1 for Meta-learning PINN loss functions

Figure 2 for Meta-learning PINN loss functions

Figure 3 for Meta-learning PINN loss functions

Figure 4 for Meta-learning PINN loss functions

Access Paper or Ask Questions

Understanding Dynamics of Nonlinear Representation Learning and Its Application

Jun 28, 2021
Kenji Kawaguchi, Linjun Zhang, Zhun Deng

Figure 1 for Understanding Dynamics of Nonlinear Representation Learning and Its Application

Figure 2 for Understanding Dynamics of Nonlinear Representation Learning and Its Application

Figure 3 for Understanding Dynamics of Nonlinear Representation Learning and Its Application

Figure 4 for Understanding Dynamics of Nonlinear Representation Learning and Its Application

Access Paper or Ask Questions

Adversarial Training Helps Transfer Learning via Better Representations

Jun 18, 2021
Zhun Deng, Linjun Zhang, Kailas Vodrahalli, Kenji Kawaguchi, James Zou

Figure 1 for Adversarial Training Helps Transfer Learning via Better Representations

Figure 2 for Adversarial Training Helps Transfer Learning via Better Representations

Figure 3 for Adversarial Training Helps Transfer Learning via Better Representations

Figure 4 for Adversarial Training Helps Transfer Learning via Better Representations

Access Paper or Ask Questions

MemStream: Memory-Based Anomaly Detection in Multi-Aspect Streams with Concept Drift

Jun 07, 2021
Siddharth Bhatia, Arjit Jain, Shivin Srivastava, Kenji Kawaguchi, Bryan Hooi

Figure 1 for MemStream: Memory-Based Anomaly Detection in Multi-Aspect Streams with Concept Drift

Figure 2 for MemStream: Memory-Based Anomaly Detection in Multi-Aspect Streams with Concept Drift

Figure 3 for MemStream: Memory-Based Anomaly Detection in Multi-Aspect Streams with Concept Drift

Figure 4 for MemStream: Memory-Based Anomaly Detection in Multi-Aspect Streams with Concept Drift

Access Paper or Ask Questions

Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth

May 26, 2021
Keyulu Xu, Mozhi Zhang, Stefanie Jegelka, Kenji Kawaguchi

Figure 1 for Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth

Figure 2 for Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth

Figure 3 for Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth

Figure 4 for Optimization of Graph Neural Networks: Implicit Acceleration by Skip Connections and More Depth

Access Paper or Ask Questions

Deep Kronecker neural networks: A general framework for neural networks with adaptive activation functions

May 20, 2021
Ameya D. Jagtap, Yeonjong Shin, Kenji Kawaguchi, George Em Karniadakis

Figure 1 for Deep Kronecker neural networks: A general framework for neural networks with adaptive activation functions

Figure 2 for Deep Kronecker neural networks: A general framework for neural networks with adaptive activation functions

Figure 3 for Deep Kronecker neural networks: A general framework for neural networks with adaptive activation functions

Figure 4 for Deep Kronecker neural networks: A general framework for neural networks with adaptive activation functions

* 26 pages, 13 figures

Access Paper or Ask Questions

A Recipe for Global Convergence Guarantee in Deep Neural Networks

Apr 15, 2021
Kenji Kawaguchi, Qingyun Sun

Figure 1 for A Recipe for Global Convergence Guarantee in Deep Neural Networks

Figure 2 for A Recipe for Global Convergence Guarantee in Deep Neural Networks

Figure 3 for A Recipe for Global Convergence Guarantee in Deep Neural Networks

Figure 4 for A Recipe for Global Convergence Guarantee in Deep Neural Networks

* Published in AAAI 2021

Access Paper or Ask Questions

On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

Feb 18, 2021
Kenji Kawaguchi

Figure 1 for On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

Figure 2 for On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

Figure 3 for On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

Figure 4 for On the Theory of Implicit Deep Learning: Global Convergence with Implicit Layers

* ICLR 2021. Selected for ICLR Spotlight (top 6% submissions)

Access Paper or Ask Questions

A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

Feb 15, 2021
Kenji Kawaguchi

Figure 1 for A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

Figure 2 for A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

Figure 3 for A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

Figure 4 for A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

* ICLR 2021. Selected for ICLR Spotlight (top 6% submissions)

Access Paper or Ask Questions

When and How Mixup Improves Calibration

Feb 11, 2021
Linjun Zhang, Zhun Deng, Kenji Kawaguchi, James Zou

Figure 1 for When and How Mixup Improves Calibration

Figure 2 for When and How Mixup Improves Calibration

Figure 3 for When and How Mixup Improves Calibration

Figure 4 for When and How Mixup Improves Calibration

Access Paper or Ask Questions

Towards Domain-Agnostic Contrastive Learning

Nov 09, 2020
Vikas Verma, Minh-Thang Luong, Kenji Kawaguchi, Hieu Pham, Quoc V. Le

Figure 1 for Towards Domain-Agnostic Contrastive Learning

Figure 2 for Towards Domain-Agnostic Contrastive Learning

Figure 3 for Towards Domain-Agnostic Contrastive Learning

Figure 4 for Towards Domain-Agnostic Contrastive Learning

Access Paper or Ask Questions

Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time

Oct 14, 2020
Ferran Alet, Kenji Kawaguchi, Maria Bauza, Nurullah Giray Kuru, Tomas Lozano-Perez, Leslie Pack Kaelbling

Figure 1 for Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time

Figure 2 for Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time

Figure 3 for Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time

Access Paper or Ask Questions

How Does Mixup Help With Robustness and Generalization?

Oct 09, 2020
Linjun Zhang, Zhun Deng, Kenji Kawaguchi, Amirata Ghorbani, James Zou

Figure 1 for How Does Mixup Help With Robustness and Generalization?

Figure 2 for How Does Mixup Help With Robustness and Generalization?

Figure 3 for How Does Mixup Help With Robustness and Generalization?

Figure 4 for How Does Mixup Help With Robustness and Generalization?

Access Paper or Ask Questions

Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks

Oct 18, 2019
Ameya D. Jagtap, Kenji Kawaguchi, George Em Karniadakis

Figure 1 for Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks

Figure 2 for Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks

Figure 3 for Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks

Figure 4 for Locally adaptive activation functions with slope recovery term for deep and physics-informed neural networks

* 19 pages, 13 figures

Access Paper or Ask Questions

Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes

Aug 05, 2019
Kenji Kawaguchi, Jiaoyang Huang

Figure 1 for Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes

Figure 2 for Gradient Descent Finds Global Minima for Generalizable Deep Neural Networks of Practical Sizes

* Accepted to appear in IEEE Proc. of the 57th Allerton Conference, UIUC, Illinois, USA

Access Paper or Ask Questions

A Stochastic First-Order Method for Ordered Empirical Risk Minimization

Jul 09, 2019
Kenji Kawaguchi, Haihao Lu

Figure 1 for A Stochastic First-Order Method for Ordered Empirical Risk Minimization

Figure 2 for A Stochastic First-Order Method for Ordered Empirical Risk Minimization

Figure 3 for A Stochastic First-Order Method for Ordered Empirical Risk Minimization

Figure 4 for A Stochastic First-Order Method for Ordered Empirical Risk Minimization

* code available at: https://github.com/kenjikawaguchi/qSGD

Access Paper or Ask Questions

Every Local Minimum is a Global Minimum of an Induced Model

Apr 07, 2019
Kenji Kawaguchi, Leslie Pack Kaelbling

Figure 1 for Every Local Minimum is a Global Minimum of an Induced Model

Figure 2 for Every Local Minimum is a Global Minimum of an Induced Model

Figure 3 for Every Local Minimum is a Global Minimum of an Induced Model

Figure 4 for Every Local Minimum is a Global Minimum of an Induced Model

Access Paper or Ask Questions

Eliminating all bad Local Minima from Loss Landscapes without even adding an Extra Unit

Jan 12, 2019
Jascha Sohl-Dickstein, Kenji Kawaguchi

Figure 1 for Eliminating all bad Local Minima from Loss Landscapes without even adding an Extra Unit

Access Paper or Ask Questions

Elimination of All Bad Local Minima in Deep Learning

Jan 02, 2019
Kenji Kawaguchi, Leslie Pack Kaelbling

Figure 1 for Elimination of All Bad Local Minima in Deep Learning

Figure 2 for Elimination of All Bad Local Minima in Deep Learning

Access Paper or Ask Questions

Effect of Depth and Width on Local Minima in Deep Learning

Nov 20, 2018
Kenji Kawaguchi, Jiaoyang Huang, Leslie Pack Kaelbling

Access Paper or Ask Questions

Depth with Nonlinearity Creates No Bad Local Minima in ResNets

Oct 21, 2018
Kenji Kawaguchi, Yoshua Bengio

Access Paper or Ask Questions