Few-shot Domain Adaptation by Causal Mechanism Transfer

Authors: Takeshi Teshima, Issei Sato, Masashi Sugiyama

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally demonstrate the effectiveness of the proposed algorithm (Section 6).
Researcher Affiliation Academia 1The University of Tokyo, Tokyo, Japan 2RIKEN, Tokyo, Japan.
Pseudocode Yes Algorithm 1 Proposed method: mechanism transfer
Open Source Code No The paper does not provide an explicit statement or link for the open-source code of the methodology described.
Open Datasets Yes We use the gasoline consumption data (Greene, 2012, p.284, Example 9.5), which is a panel data of gasoline usage in 18 of the OECD countries over 19 years.
Dataset Splits Yes To perform hyperparameter selection as well as early-stopping, we record the leave-one-out cross-validation (LOOCV) mean-squared error on the target training data every 20 epochs and select its minimizer.
Hardware Specification Yes All experiments were conducted on an Intel Xeon(R) 2.60 GHz CPU with 132 GB memory.
Software Dependencies No The paper mentions 'Python using the Py Torch library (Paszke et al., 2019) or the R language (R Core Team, 2018)' but does not provide specific version numbers for these software components.
Experiment Setup Yes We use the Adam optimizer (Kingma & Ba, 2017) with fixed parameters (β1, β2, ϵ) = (0.9, 0.999, 10 8), fixed initial learning rate 10 3, and the maximum number of epochs 300. The number of hidden units for ψd is chosen from {10, 20} and the coefficient of weightdecay from 10{ 2, 1}. The ℓ2 regularization coefficient λ of KRR is chosen from λ 2{ 10,...,10} following Cortes et al. (2019).