Theory

Ground-Truth or DAER: Selective Re-Query of Secondary Information ICCV 2021 Group testing and local search: is there a computational-statistical gap? COLT 2021 Heating up decision boundaries: isocapacitory saturation, adversarial scenarios and generalization bounds ICLR 2021 Heavy Tails in SGD and Compressibility of Overparametrized Neural Networks NeurIPS 2021 High-probability Bounds for Non-Convex Stochastic Optimization with Heavy Tails NeurIPS 2021 Hilbert Sinkhorn Divergence for Optimal Transport CVPR 2021 Hoeffding's Inequality for General Markov Chains and Its Applications to Statistical Learning JMLR 2021 How Does Fine-tuning Affect the Geometry of Embedding Space: A Case Study on Isotropy EMNLP 2021 How Does Loss Function Affect Generalization Performance of Deep Learning? Application to Human Age Estimation ICML 2021 How Good (really) are Grammatical Error Correction Systems? EACL 2021 How Hard to Tell? Complexity of Belief Manipulation Through Propositional Announcements IJCAI 2021 How Many Layers and Why? An Analysis of the Model Depth in Transformers ACL 2021 How rotational invariance of common kernels prevents generalization in high dimensions ICML 2021 How Tight Can PAC-Bayes be in the Small Data Regime? NeurIPS 2021 How to Learn when Data Reacts to Your Model: Performative Gradient Descent ICML 2021 How Well Generative Adversarial Networks Learn Distributions JMLR 2021 Hyperparameter Optimization Is Deceiving Us, and How to Stop It NeurIPS 2021 Identity-aware Graph Neural Networks AAAI 2021 Implicit Bias of SGD for Diagonal Linear Networks: a Provable Benefit of Stochasticity NeurIPS 2021 Implicit Convex Regularizers of CNN Architectures: Convex Optimization of Two- and Three-Layer Networks in Polynomial Time ICLR 2021 Implicit Regularization in Matrix Sensing via Mirror Descent NeurIPS 2021 Implicit Regularization in ReLU Networks with the Square Loss COLT 2021 Implicit Regularization in Tensor Factorization ICML 2021 Implicit Regularization via Neural Feature Alignment AISTATS 2021 Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning ICLR 2021