Mediated Uncoupled Learning: Learning Functions without Direct Input-output Correspondences

Authors: Ikko Yamane, Junya Honda, Florian Yger, Masashi Sugiyama

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We prove statistical consistency and error bounds of our method and experimentally confirm its practical usefulness.
Researcher Affiliation Academia 1LAMSADE, CNRS, Université Paris-Dauphine, PSL Research University, 75016 PARIS, FRANCE 2RIKEN AIP, Tokyo, Japan 3Kyoto University, Kyoto, Japan 4The University of Tokyo, Tokyo, Japan. Correspondence to: Ikko Yamane <ikko.yamane@dauphine.psl.eu>.
Pseudocode Yes Algorithm 1 Two-Step Regressed Regression (2Step-RR); Algorithm 2 Joint Regressed Regression (Joint-RR)
Open Source Code Yes The code will be available on https://github.com/i-yamane/mediated_uncoupled_learning.
Open Datasets Yes MNIST (Le Cun et al., 1994), Fashion-MNIST (Xiao et al., 2017), CIFAR-10, and CIFAR-100 (Krizhevsky et al., 2009).
Dataset Splits Yes We use 1,000 mediated uncoupled data for training and 10,000 coupled (X, Y )-data for test evaluation. (...) We use randomly sampled 10,000 mediated uncoupled data for training and 10,000 coupled (X, Y )-data for test evaluation.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory specifications) used for running the experiments.
Software Dependencies No The paper mentions 'Py Torch (Paszke et al., 2019)' and 'Adam (Kingma & Ba, 2017)' but does not provide specific version numbers for PyTorch or any other libraries or dependencies used.
Experiment Setup Yes We train all models with Adam (Kingma & Ba, 2017) for 200 epochs. (...) We use the default values of the implementation provided by Py Torch (Paszke et al., 2019) for all the parameters of Adam: the learning rate is 0.001, and β is (0.9, 0.999). (...) We turn off the weight decay and set the other tuning parameters of Adam as in Py Torch (Paszke et al., 2019): the learning rate is 0.001, the β is (0.9, 0.999).