Research map / Learning theory and algorithms / Prediction and decision losses
Using Shapley Values and Variational Autoencoders to Explain Predictive Models with Dependent Mixed Features
Lars H. B. Olsen, Ingrid K. Glad, Martin Jullum, Kjersti Aas
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use VAEAC instead of Gaussian approach | Shapley value estimation · Burr continuous, M=5, N_train=1000 | EC1: 0.0424 -> 0.0264 | Improved |
| increase training size from 1000 to 5000 | VAEAC · Burr continuous, M=10 | EC1: 0.0288 -> 0.0218 | Improved |
| increase training size from 1000 to 5000 | Independence approach · Burr continuous, M=10 | EC1: 0.1082 -> 0.1160 | Hurt |
| increase training size from 1000 to 5000 | Gaussian approach · Burr continuous, M=10 | EC1: 0.0384 -> 0.0384 | No clear effect |
| use coalition-focused masking (VAEAC_C) instead of uniform masking | VAEAC · Burr continuous, M=100, N_train=10000 | EC1: 0.1013 -> 0.0504 | Improved |
| use coalition-focused masking (VAEAC_C) instead of uniform masking | VAEAC · Burr continuous, M=250, N_train=25000 | EC1: 0.6891 -> 0.1474 | Improved |
| use coalition-focused masking (VAEAC_C) instead of uniform masking | VAEAC · Burr continuous, M=25, N_train=1000 | EC1: 0.0321 -> 0.0346 | Hurt |
| use VAEAC_C instead of Gaussian at limited training size | Shapley estimation · Burr continuous, M=100, N_train=5000 | EC1: 0.0555 -> 0.0594 | Hurt |
| use VAEAC instead of independence under strong dependence | Shapley estimation · Mixed M=4, N_train=5000, rho=0.9 | EC1: 0.6464 -> 0.0489 | Improved |
| use VAEAC instead of independence for independent features | Shapley estimation · Mixed M=4, N_train=5000, rho=0.0 | EC1: 0.0240 -> 0.0523 | Hurt |
| use VAEAC instead of ctree with moderate training set | Shapley estimation · Mixed M=4, N_train=1000, rho=0.1 | EC1: 0.0706 -> 0.0770 | Hurt |
| use VAEAC instead of ctree for larger categorical dimension | Shapley estimation · Categorical M=7 L=4, rho=0.9 | EC1: 0.0736 -> 0.0647 | Improved |
Closest papers by meaning
- Shapley Residuals: Quantifying the limits of the Shapley value for explanations (NeurIPS 2021 · similarity 0.89)
- shapiq: Shapley Interactions for Machine Learning (NeurIPS 2024 · similarity 0.88)
- Unlocking the Game: Estimating Games in Möbius Representation for Explanation and High-Order Interaction Detection (AAAI 2025 · similarity 0.86)
- Provably Accurate Shapley Value Estimation via Leverage Score Sampling (ICLR 2025 · similarity 0.86)
- InstaSHAP: Interpretable Additive Models Explain Shapley Values Instantly (ICLR 2025 · similarity 0.86)
- Support Vector-based Estimation of Multilinear Games for Feature Selection and Explanation (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
