Regularized Learning for Domain Adaptation under Label Shifts

Authors: Kamyar Azizzadenesheli, Anqi Liu, Fanny Yang, Animashree Anandkumar

ICLR 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on the CIFAR-10 and MNIST datasets show that RLLS improves classification accuracy, especially in the low sample and large-shift regimes, compared to previous methods.
Researcher Affiliation Academia Kamyar Azizzadenesheli University of California, Irvine kazizzad@uci.edu Anqi Liu California Institute of Technology anqiliu@caltech.edu Fanny Yang Institute of Theoretical Studies, ETH Zürich fan.yang@stat.math.ethz.ch Animashree Anandkumar California Institute of Technology anima@caltech.edu
Pseudocode Yes Algorithm 1 Regularized Learning of Label Shift (RLLS)
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available.
Open Datasets Yes Experiments on the CIFAR-10 and MNIST datasets... MNIST (Le Cun & Cortes, 2010) and CIFAR10 (Krizhevsky & Hinton, 2009) datasets.
Dataset Splits No The paper describes dividing the source data into subsets for weight estimation and classifier training (e.g., "divide it into two sets where we use (1 β)np samples in set Dweight p to compute the estimate bw and the remaining n = βnp in the set Dclass p to find the classifier"). It also mentions that a hyperparameter 'can be chosen using standard cross validation methods'. However, it does not explicitly specify a conventional train/validation/test dataset split percentages or sample counts used for model selection or hyperparameter tuning in their reported experiments.
Hardware Specification No The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper mentions using a 'two-layer fully connected neural network for MNIST and a Res Net-18 (He et al., 2016) for CIFAR10,' but does not specify any software dependencies or their version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes In order to compute bω = 1+bθ in Eq. (3), we call a built-in solver to directly solve the low dimensional problem minθ b Cθ bb 2 + C θ 2 where we empirically observer that 0.01 times of the true C yields in a better estimator on various levels of label shift pre-computed beforehand. It is worth noting that 0.001 makes the theoretical bound in Lemma. 1 O(1/0.01) times bigger. We thus treat it as a hyperparameter that can be chosen using standard cross validation methods. Finally, we train a classifier on the source samples weighted by bω, where we use a two-layer fully connected neural network for MNIST and a Res Net-18 (He et al., 2016) for CIFAR10.