Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Efficient Transformed Gaussian Processes for Non-Stationary Dependent Multi-class Classification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Transform one shared GP sample into class-specific latent functions with input-dependent flows | ETGP · characterfont | accuracy: 0.29 -> 0.36 | Helped |
| Transform one shared GP sample into class-specific latent functions with input-dependent flows | ETGP · devangari | accuracy: 0.93 -> 0.96 | Helped |
| Transform one shared GP sample into class-specific latent functions with input-dependent flows | ETGP · avila | accuracy: SVGP: 0.962; TANH: 0.985; LINEAR: 0.970; RBFCORR: 0.988 | Mixed |
| Transform one shared GP sample into class-specific latent functions with input-dependent flows | ETGP · absenteeism | accuracy: SVGP: 0.314; RBFCORR: 0.319; LINEAR: 0.307; TANH: 0.315 | No clear effect |
| Share kernels and inducing points across classes rather than use class-specific SVGP settings (ablation) | multi-class SVGP · characterfont and devangari | accuracy | Hurt |
| Use fewer inducing points (50 or 100) (ablation) | ETGP · vowel | accuracy | Helped |
| Use fewer inducing points (50 or 100) (ablation) | ETGP · avila and absenteeism | accuracy | No clear effect |
| Use SAL flows instead of LINEAR, TANH, or other ETGP flow choices (ablation) | ETGP · all five UCI datasets | accuracy | Hurt |
| Use Bayesian rather than point-estimate flow-parameter predictions (ablation) | ETGP · small datasets | log-likelihood | Helped |
| Use Bayesian rather than point-estimate flow-parameter predictions (ablation) | ETGP · medium and large datasets | accuracy | No clear effect |
| Use an input-dependent flow rather than a non-stationary NNET or ARCCOS kernel | ETGP · UCI classification datasets | accuracy | Helped |
| Use ETGP rather than SVGP | multi-class classification · UCI classification datasets | training time: around one order of magnitude | Helped |
Closest papers by meaning
- Non-Gaussian Gaussian Processes for Few-Shot Regression (NeurIPS 2021 · similarity 0.87)
- Scalable Multi-Class Gaussian Process Classification using Expectation Propagation (ICML 2017 · similarity 0.87)
- GP-Tree: A Gaussian Process Classifier for Few-Shot Incremental Learning (ICML 2021 · similarity 0.86)
- Random Feature Expansions for Deep Gaussian Processes (ICML 2017 · similarity 0.86)
- Robust and Conjugate Gaussian Process Regression (ICML 2024 · similarity 0.85)
- Domain Invariant Learning for Gaussian Processes and Bayesian Exploration (AAAI 2024 · similarity 0.84)
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.