Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Minimizing Negative Transfer of Knowledge in Multivariate Gaussian Processes: A Scalable and Regularized Approach
Raed Al Kontar, Garvesh Raskutti, Shiyu Zhou
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 |
|---|---|---|---|
| add penalized pairwise shared latent function (MGCP-RD) instead of full MGCP | multivariate GP · Exchange rates | MAE: 0.203 -> 0.156 | Improved |
| add pairwise regularized decomposition instead of full MGCP | multivariate GP · Exchange rates | SMSE: 1.666 -> 1.293 | Improved |
| average parameter estimates across pairs (MGCP-P) instead of averaging predictions (ablation) | pairwise MGP · Exchange rates | MAE: 0.156 -> 0.952 | Hurt |
| use inducing variables (MGCP-I) instead of regularized pairwise decomposition (ablation) | multivariate GP · Exchange rates | MAE: 0.156 -> 0.218 | Hurt |
| add penalization of shared latent function when outputs lack commonalities (RD vs full MGCP) | MGCP · Setting I no negative transfer (N=5) | prediction error | No clear effect |
| analyze negatively transferring outputs separately (MGCP-Sep) instead of jointly (ablation) | MGCP · Setting III simulation | prediction error | Improved |
| use MGCP-RD instead of MGCP-Sep | MGCP-Sep · Setting III simulation | prediction error | Improved |
| increase number of outputs N in MGCP-RD (ablation) | MGCP-RD · Setting II simulation | prediction error | Improved |
| use pairwise decomposition: build N-1 models instead of N(N-1)/2 | MGCP-RD vs MGCP-P · Setting II (N=50) | model building time: 104 minutes -> 6.8 minutes | Improved |
Closest papers by meaning
- Regularized Multi-Output Gaussian Convolution Process With Domain Adaptation (TPAMI 2022 · similarity 0.90)
- Convolutional Gaussian Processes (NeurIPS 2017 · similarity 0.88)
- Generic Inference in Latent Gaussian Process Models (JMLR 2019 · similarity 0.88)
- Heterogeneous Multi-output Gaussian Process Prediction (NeurIPS 2018 · similarity 0.87)
- Partial Information Decomposition via Normalizing Flows in Latent Gaussian Distributions (NeurIPS 2025 · similarity 0.86)
- Turbocharging Gaussian Process Inference with Approximate Sketch-and-Project (NeurIPS 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.
