Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Stochastic Variational Deep Kernel Learning
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 |
|---|---|---|---|
| joint stochastic variational deep kernel learning (SV-DKL) instead of stand-alone DNN | DNN on airline delay · Airline delay (5.9M points) | accuracy: +0.008 (0.773 to 0.781) | Helped |
| joint SV-DKL instead of stand-alone DNN | DNN on UCI classification · Adult, Connect4, Diabetes, Covtype | accuracy: +0.005 to +0.036 | Helped |
| joint SV-DKL instead of stand-alone CNN | CNN on image benchmarks · MNIST, CIFAR10, SVHN, MNIST-Binary | accuracy: +0.0001 to +0.0112 | Helped |
| adding an extra Q x c hidden layer (DNN+) | DNN · Airline delay | accuracy: 0.773 to 0.722 | Hurt |
| adding an extra Q x c hidden layer (CNN+) | CNN · MNIST, CIFAR10, SVHN | accuracy: small changes both ways | Mixed |
| GP on pre-trained DNN features (DNN+GP) instead of stand-alone DNN | DNN · Airline delay and UCI | accuracy: airline +0.0019; UCI Connect4 +0.006 but Adult +0.001 and others mixed | Mixed |
| KISS-GP based variational SV-DKL instead of KLSP-GP | GP classifier · Airline delay | accuracy: ~0.675 to 0.781 | Helped |
| ResNet-20 base replaced with SV-DKL joint training | ResNet-20 · CIFAR10 | accuracy: 0.901 to 0.910 | Helped |
| joint SV-DKL on AlexNet variant | AlexNet variant · ImageNet (20 categories) | accuracy: 0.6877 to 0.7067 | Helped |
| structure-exploiting Kronecker/Toeplitz variational sampler vs standard O(m^3) methods (ablation) | SV-DKL inference · Airline delay scalability (runtime vs inducing points) | runtime: runtime grows slowly for m<2000 vs others | Helped |
| number of MC samples T=1 for expectation estimation | SV-DKL · Training | accuracy | No clear effect |
Closest papers by meaning
- Stochastic Kernel Regularisation Improves Generalisation in Deep Kernel Machines (NeurIPS 2024 · similarity 0.90)
- Kernel Implicit Variational Inference (ICLR 2018 · similarity 0.90)
- Deep Attentive Variational Inference (ICLR 2022 · similarity 0.89)
- Amortized Variational Deep Kernel Learning (ICML 2024 · similarity 0.88)
- VIKING: Deep variational inference with stochastic projections (NeurIPS 2025 · similarity 0.87)
- Convolutional Deep Kernel Machines (ICLR 2024 · similarity 0.87)
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.