Distributionally Robust Post-hoc Classifiers under Prior Shifts

Authors: Jiaheng Wei, Harikrishna Narasimhan, Ehsan Amid, Wen-Sheng Chu, Yang Liu, Abhishek Kumar

ICLR 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we empirically demonstrate the effectiveness of our proposed method DROPS, for the tasks of class-imbalanced learning and group distributional robustness.
Researcher Affiliation Collaboration Jiaheng Wei UC Santa Cruz Harikrishna Narasimhan Google Research Ehsan Amid Google Research Wen-Sheng Chu Google Research Yang Liu UC Santa Cruz Abhishek Kumar Google Research Work done during an internship at Google Research, Brain Team.
Pseudocode No The paper describes algorithmic steps in prose (Section 4.3, 'Step 1: updating λ(t).', 'Step 2: updating g(t).', 'Step 3: scaling the predictions.') but does not include formal pseudocode or algorithm blocks.
Open Source Code Yes An empirical implementation is available at https://github.com/weijiaheng/Drops.
Open Datasets Yes For CIFAR-10, and CIFAR-100 datasets, we down-sample the number of samples for each class to simulate the class-imbalance as done in earlier works (Cui et al., 2019; Cao et al., 2019).
Dataset Splits Yes A balanced held-out validation set is utilized for hyper-parameter tuning. All baseline models are picked by referring to the δ = 1.0-worst case performances on the validation set, which is made up of the last 10% of the original CIFAR training dataset.
Hardware Specification No The paper mentions using a specific model architecture ('Pre Act Res Net 18') but does not provide any details about the hardware (e.g., GPU models, CPU, or memory) used for the experiments.
Software Dependencies No The paper mentions using 'SGD optimizer' and specific learning rate schedules, but does not provide version numbers for any software libraries, frameworks (e.g., TensorFlow, PyTorch), or programming languages used.
Experiment Setup Yes All methods are trained with the same architecture (Pre Act Res Net 18 (He et al., 2016)) with 5 random seeds, same data augmentation techniques, the same SGD optimizer with a momentum of 0.9 with Nesterov acceleration. All methods share the same initial learning rate of 0.1 and a piece-wise constant learning rate schedule of [10 2, 10 3, 10 4] at [30, 80, 110] epochs, respectively. We use a batch size of 128 for all methods and train the model for 140 epochs.