Active Invariant Causal Prediction: Experiment Selection through Stability
Authors: Juan L. Gamella, Christina Heinze-Deml
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, we analyze the performance of the proposed policies in both population and finite-regime experiments. We evaluate policies that use different combinations of the strategies in both the population and finite sample setting, using simulated data from randomly chosen linear SCMs. |
| Researcher Affiliation | Academia | Juan L. Gamella Seminar for Statistics ETH Zurich Switzerland gajuan@ethz.ch Christina Heinze-Deml Seminar for Statistics ETH Zurich Switzerland heinzedeml@stat.math.ethz.ch |
| Pseudocode | Yes | Algorithm 1: A-ICP |
| Open Source Code | Yes | Further details about the experimental settings and links to code to reproduce the experiments can be found in Appendix E. |
| Open Datasets | No | The paper states that it uses "simulated data from randomly chosen linear SCMs" but does not provide concrete access information (e.g., link, DOI, citation) for a publicly available dataset. |
| Dataset Splits | No | The paper mentions "picking the regularization parameter for each SCM by cross validation" which implies a validation process, but it does not provide specific details on overall dataset splits for training, validation, or testing (e.g., exact percentages or sample counts for each split). |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using "the Lasso [36]" which refers to a method, but it does not specify any software libraries or packages with version numbers (e.g., "scikit-learn 0.24") required for replication. |
| Experiment Setup | Yes | For the sample allocation, we fix the size of the sample collected per intervention; we perform experiments for 10, 100 and 1000 observations per sample. The total number of iterations is 50. Here, α = 0.01. Here, α = 0.1. |