ICLR2022 Interesting Papers

原文: http://zhuanlan.zhihu.com/p/499010769

  • When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
  • Self-supervised Learning is More Robust to Dataset Imbalance
  • Looking Back on Learned Experiences For Class/task Incremental Learning
  • Path Auxiliary Proposal for MCMC in Discrete Space
  • On Bridging Generic and Personalized Federated Learning for Image Classification
  • Improved deterministic l2 robustness on CIFAR-10 and CIFAR-100
  • Learning Strides in Convolutional Neural Networks
  • How to Robustify Black-Box ML Models? A Zeroth-Order Optimization Perspective
  • NASPY: Automated Extraction of Automated Machine Learning Models
  • Scalable Sampling for Nonsymmetric Determinantal Point Processes
  • Strength of Minibatch Noise in SGD
  • A Fine-Grained Analysis on Distribution Shift
  • Minibatch vs Local SGD with Shuffling: Tight Convergence Bounds and Beyond
  • Continual Learning with Recursive Gradient Optimization
  • Learning meta-features for AutoML
  • Scalable One-Pass Optimisation of High-Dimensional Weight-Update Hyperparameters by Implicit Differentiation
  • Exploring the Limits of Large Scale Pre-training
  • SUMNAS: Supernet with Unbiased Meta-Features for Neural Architecture Search
  • On Redundancy and Diversity in Cell-based Neural Architecture Search
  • NASI: Label- and Data-agnostic Neural Architecture Search at Initialization
  • NASViT: Neural Architecture Search for Efficient Vision Transformers with Gradient Conflict aware Supernet Training
  • Surrogate NAS Benchmarks: Going Beyond the Limited Search Spaces of Tabular NAS Benchmarks
  • Automatic Loss Function Search for Predict-Then-Optimize Problems with Strong Ranking Property
Flag Counter