A General Class of Transfer Learning Regression without Implementation Cost
Authors: Shunya Minami, Song Liu, Stephen Wu, Kenji Fukumizu, Ryo Yoshida8992-8999
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate its simplicity, generality, and applicability using various real data applications. |
| Researcher Affiliation | Academia | 1 The Graduate University for Advanced Studies (SOKENDAI) 2 University of Bristol 3 The Institute of Statistical Mathematics 4 National Institute for Materials Science |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions using 'standard libraries of the R language (glmnet, ranger, and MXNet)' but does not provide concrete access to source code for the methodology described in this paper. |
| Open Datasets | Yes | The proposed method was applied to five real data analyses in materials science and robotics applications: (i) multiple properties of organic polymers and inorganic compounds (Yamada et al. 2019), (ii) multiple properties of polymers (Kim et al. 2018) and lowmolecular-weight compounds (monomers, unpublished data), (iii) properties of donor molecules in organic solar cells (Paul et al. 2019) obtained from experiments (Lopez et al. 2016) and quantum chemical calculations (Pyzer-Knapp, Li, and Aspuru-Guzik 2015), (iv) formation energies of various inorganic compounds and crystal polymorphisms of Si O2 and Cd I2 (Jain et al. 2013), and (v) the feed-forward torques required to follow a desired trajectory at seven joints of a SARCOS anthropomorphic robot arm (Williams and Rasmussen 2006). |
| Dataset Splits | Yes | We choose the best model based on the 5-fold cross validation. The resulting model was used to predict all the remaining data, and the MSE was evaluated. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper states 'We simply used the standard libraries of the R language (glmnet, ranger, and MXNet)' but does not provide specific version numbers for these software components. |
| Experiment Setup | Yes | In the source task, the entire dataset was used to train fs(x) under default settings of software packages without adjusting hyperparameters. In all cases, 50 randomly selected samples were used to train fθw(x). We used the linear ridge regression to estimate fθw with the hyperparameter on the ℓ2-regularization that was fixed at λ = 0.0001. |