Linear tree shap
Authors: peng yu, Albert Bifet, Jesse Read, Chao Xu
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We run an experiment on both regression dataset adult and classification dataset conductor(summary in Table 2) to compare both our method and two popular algorithms, Tree Shap and Fast Tree Shap. We explain Trees with depths ranging from 2 to 18. And to align the performance across different depths of trees, we plot the ratio between the time of Tree Shap and the time of all methods in Figure 5. We run every algorithm on the same test set 5 times to get both average speeds up and the error bar. And for fair comparison purposes, all methods are limited to using a single core. |
| Researcher Affiliation | Collaboration | Peng Yu1,2,5 Chao Xu1 Albert Bifet2,4 Jesse Read3 1 University of Electronic Science and Technology of China 2 LTCI, T el ecom Paris, IP Paris 3 LIX, Ecole Polytechnique, IP Paris 4 AI Institute, University of Waikato 5 Shopify |
| Pseudocode | Yes | COMPUTESUMMARYPOLYNOMIALS(x, v, C): ... Algorithm 2: Obtain the summary polynomial for each node. AGGREGATESHAPLEY(x, v, G): ... Algorithm 3: Obtain the Shapley value vector. LINEARTREESHAP(x, Tf): ... Algorithm 4: The entire LINEARTREESHAP algorithm. |
| Open Source Code | Yes | 3. If you ran experiments... (a) Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] |
| Open Datasets | Yes | We run an experiment on both regression dataset adult and classification dataset conductor(summary in Table 2). Datasets: adult [4] 48,842 64 Classification 2; conductor [3] 21,263 81 Regression. |
| Dataset Splits | No | The paper mentions running algorithms on a 'test set' but does not explicitly provide details about a validation set or specific training/validation/test splits by percentages or counts. |
| Hardware Specification | No | The paper states 'all methods are limited to using a single core' but does not specify any particular CPU model, GPU, memory, or other detailed hardware specifications for the experimental setup. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x) that are needed to replicate the experiments. |
| Experiment Setup | Yes | We explain Trees with depths ranging from 2 to 18. And for fair comparison purposes, all methods are limited to using a single core. |