Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Scalable Computation of Causal Bounds
Authors: Madhumitha Shridharan, Garud Iyengar
JMLR 2023 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we report the results of our numerical experiments validating and extending the benefits from the new methods proposed in Sections 3 and 5. In Table 2 we report the results of the experiments for constructing the pruned LPs (5) and (7). The results clearly show the significant runtime improvement provided by our method compared to the benchmark on the five examples. In Figure 5 we plot the upper and lower bounds computed by solving (25) as a function of δ for 10 randomly generated instances of p in Example A. We tested Algorithm 5 on 100 instances of each of the examples in Appendix B for which bounds were not available in closed form. |
| Researcher Affiliation | Academia | Madhumitha Shridharan EMAIL Garud Iyengar EMAIL Department of Industrial Engineering and Operations Research Columbia University New York City, NY 10027, USA |
| Pseudocode | Yes | Algorithm 1 Procedure to efficiently construct LPs (5), (7) Input: (i) causal graph G, (ii) query Q = P(VO(VI = q I) = q O|VA = q A), (iii) conditional probability distribution pv B.v A = P(VB = v B|VA = v A), for all v A, v B. Output: Pruned LPs (5) and (7) H for h : VA VB do if h is valid (Theorem 5) then H H {h} Compute c L h using Theorem 6 Compute c U h using Theorem 7 return LPs (5) and (7) constructed using (H, c L, c U) |
| Open Source Code | No | The paper does not provide any explicit statement about the release of source code, nor does it include a link to a code repository or reference code in supplementary materials. |
| Open Datasets | No | The paper describes data generating processes for its examples, indicating the data was simulated or generated by the authors. For instance, in Appendix B, Example A states: 'The causal graph for this example is display in Figure 7a... and data generating process used to generate the input data is given by Z1 Bernoulli(logit 1(UA)).' No specific links, DOIs, or citations to existing publicly available datasets are provided. |
| Dataset Splits | No | The paper describes generating data instances for experiments (e.g., 'for 10 randomly generated instances of p in Example A', 'on 100 instances of each of the examples in Appendix B'). This implies data generation for each instance rather than the use of predefined dataset splits for training, validation, or testing. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments. There are no mentions of GPU or CPU models, memory specifications, or cloud computing instance types. |
| Software Dependencies | No | The paper does not list any specific software dependencies with version numbers, such as programming languages, libraries, or solvers used for implementation. |
| Experiment Setup | No | The paper describes algorithmic procedures and their computational performance, but it does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs), model initialization, or training schedules. |