Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Inference with Tail-adaptive f-Divergence
Dilin Wang, Hao Liu, Qiang Liu
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 tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Boston | test RMSE: 2.956±0.171 -> 2.828±0.177 | Improved |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.5 | Bayesian neural network regression · Concrete | test RMSE: 5.381±0.111 -> 5.371±0.115 | No clear effect |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Energy | test log-likelihood: -1.795±0.029 -> -1.758±0.034 | Improved |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.5 | Bayesian neural network regression · Kin8nm | test RMSE: 0.083±0.001 -> 0.085±0.001 | Hurt |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Naval | test RMSE: 0.001±0.000 -> 0.001±0.000 | No clear effect |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.5 | Bayesian neural network regression · Combined | test RMSE: 4.154±0.042 -> 4.116±0.032 | Improved |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Wine | test RMSE: 0.634±0.007 -> 0.636±0.008 | Hurt |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Yacht | test log-likelihood: -1.751±0.056 -> -1.711±0.059 | Improved |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.0 | Bayesian neural network regression · Protein | test RMSE: 4.565±0.026 -> 4.487±0.019 | Improved |
| Use tail-adaptive f-divergence (β=-1.0) instead of α=0.5 | Bayesian neural network regression · Year | test log-likelihood: -3.518±0.042 -> -3.570±0.037 | Hurt |
| Use tail-adaptive f-divergence instead of fixed α-divergence | SAC · MuJoCo benchmark tasks | final average reward | Improved |
| Increase α from the original SAC setting α=0 | SAC · MuJoCo environments | average reward | Improved |
Closest papers by meaning
- f-Divergence Variational Inference (NeurIPS 2020 · similarity 0.88)
- Asymptotics of Alpha-Divergence Variational Inference Algorithms with Exponential Families (NeurIPS 2024 · similarity 0.88)
- Monotonic Alpha-divergence Minimisation for Variational Inference (JMLR 2023 · similarity 0.87)
- Generalized Variational Inference via Optimal Transport (AAAI 2024 · similarity 0.87)
- Forward $\chi^2$ Divergence Based Variational Importance Sampling (ICLR 2024 · similarity 0.86)
- Globally Convergent Variational Inference (NeurIPS 2024 · similarity 0.86)
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.
