Flow-based Attribution in Graphical Models: A Recursive Shapley Approach
Authors: Raghav Singal, George Michailidis, Hoiyi Ng
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To illustrate how RSV facilitates non-linear mediation analysis, we use the causal unfairness example (Frye et al., 2020b). ... To understand unresolved discrimination, we generate multiple datasets by varying a4 while fixing (a2, a3) at (1, 1). For each dataset, we fit a probabilistic model at nodes 3 (as a function of X1) and Y (as a function of X1, X2, X3); details in Appendix H. For each estimated model, we compute RSV by considering two applicants: (X(1) 1 , X(1) 2 ) = (0, 0) and (X(2) 1 , X(2) 2 ) = (1, 0); i.e., different value of the sensitive attribute, but same score. ... In Figure 11, we show attributions to the fair (X1 X3 Y ) and the unfair (X1 Y ) channels as a function of a4. |
| Researcher Affiliation | Collaboration | 1Amazon 2University of Florida. Correspondence to: RS <rs3566@columbia.edu>, GM <gmichail@ufl.edu>, HN <nghoiyi@amazon.com>. |
| Pseudocode | Yes | Algorithm 1 RSV(N, E) ... Algorithm 2 RSVjk(E0, . . . , En) |
| Open Source Code | No | The paper does not include any explicit statements about releasing source code or links to a code repository for the methodology described. |
| Open Datasets | No | The paper refers to generating multiple datasets for an example ('we generate multiple datasets by varying a4 while fixing (a2, a3) at (1, 1). For each dataset, we fit a probabilistic model...'), but it does not provide concrete access information (link, DOI, repository, or formal citation) for a publicly available or open dataset. |
| Dataset Splits | No | The paper does not provide specific details on how the datasets were split into training, validation, and test sets, nor does it refer to standard predefined splits or cross-validation. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments (e.g., GPU models, CPU types, memory). |
| Software Dependencies | No | The paper does not list any specific software components with their version numbers that would be necessary to replicate the experiments. |
| Experiment Setup | No | The paper describes how data was generated for Example 3 ('we generate multiple datasets by varying a4 while fixing (a2, a3) at (1, 1)') and how RSV was applied to specific inputs ('we compute RSV by considering two applicants: (X(1) 1 , X(1) 2 ) = (0, 0) and (X(2) 1 , X(2) 2 ) = (1, 0)'), but it does not provide typical experimental setup details such as hyperparameters for model training (e.g., learning rate, batch size, optimizer settings). |