Theory

The balancing principle for parameter choice in distance-regularized domain adaptation NeurIPS 2021 The Hintons in your Neural Network: a Quantum Field Theory View of Deep Learning ICML 2021 The Implicit Bias of Minima Stability: A View from Function Space NeurIPS 2021 The Lipschitz Constant of Self-Attention ICML 2021 The Neural Tangent Link Between CNN Denoisers and Non-Local Filters CVPR 2021 The Spectrum of Fisher Information of Deep Networks Achieving Dynamical Isometry AISTATS 2021 The staircase property: How hierarchical structure can guide deep learning NeurIPS 2021 Tight Bounds on the Smallest Eigenvalue of the Neural Tangent Kernel for Deep ReLU Networks ICML 2021 Topological Detection of Trojaned Neural Networks NeurIPS 2021 Topological Uncertainty: Monitoring Trained Neural Networks through Persistence of Activation Graphs IJCAI 2021 Toward Better Generalization Bounds with Locally Elastic Stability ICML 2021 Towards Sharper Generalization Bounds for Structured Prediction NeurIPS 2021 Towards Understanding Learning in Neural Networks with Linear Teachers ICML 2021 Training deep residual networks for uniform approximation guarantees L4DC 2021 Training Neural Networks is ER-complete NeurIPS 2021 Understanding How Over-Parametrization Leads to Acceleration: A case of learning a single teacher neuron ACML 2021 Understanding Model Robustness to User-generated Noisy Texts EMNLP 2021 UnICORNN: A recurrent model for learning very long time dependencies ICML 2021 Uniform Convergence, Adversarial Spheres and a Simple Remedy ICML 2021 Verifiability and Predictability: Interpreting Utilities of Network Architectures for Point Cloud Processing CVPR 2021 When Are Solutions Connected in Deep Networks? NeurIPS 2021 When does gradient descent with logistic loss interpolate using deep networks with smoothed ReLU activations? COLT 2021 Whitening and Second Order Optimization Both Make Information in the Dataset Unusable During Training, and Can Reduce or Prevent Generalization ICML 2021 Why Adversarial Interaction Creates Non-Homogeneous Patterns: A Pseudo-Reaction-Diffusion Model for Turing Instability AAAI 2021 Adversarial Training is a Form of Data-dependent Operator Norm Regularization NeurIPS 2020