Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
The Deep Weight Prior
Andrei Atanov, Arsenii Ashukha, Kirill Struminsky, Dmitriy Vetrov, Max Welling
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 a Deep Weight Prior learned from related-domain kernels instead of standard normal or log-uniform priors | Bayesian CNN variational inference · MNIST and CIFAR-10 with limited training data | test accuracy | Improved |
| Initialize fixed convolutional filters from the Deep Weight Prior instead of Xavier initialization | CNN random feature extraction · MNIST and CIFAR-10 | test accuracy | Improved |
| Initialize fixed convolutional filters with learned kernels instead of Xavier initialization | CNN random feature extraction · MNIST and CIFAR-10 | test accuracy | Improved |
| Replace Deep Weight Prior initialization with learned-filter initialization | CNN random feature extraction · MNIST and CIFAR-10 | test accuracy | Mixed |
| Initialize convolutional layers from the Deep Weight Prior instead of Xavier initialization | Convolutional VAE and CNN training · MNIST and CIFAR-10 | training speed | Improved |
| Replace Deep Weight Prior initialization with learned-filter initialization | Convolutional VAE and CNN training · MNIST and CIFAR-10 | training speed | No clear effect |
| Replace Deep Weight Prior with a learned multivariate Gaussian prior (ablation) | Bayesian CNN variational inference · MNIST | test accuracy | Hurt |
Closest papers by meaning
- Specifying Weight Priors in Bayesian Deep Neural Networks with Empirical Bayes (AAAI 2020 · similarity 0.88)
- Bayesian Neural Network Priors Revisited (ICLR 2022 · similarity 0.88)
- Posterior and Variational Inference for Deep Neural Networks with Heavy-Tailed Weights (JMLR 2025 · similarity 0.87)
- A Bayesian Approach to Data Point Selection (NeurIPS 2024 · similarity 0.85)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · similarity 0.84)
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · 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.
