Stochastic natural gradient descent draws posterior samples in function space
원문 초록
Recent work has argued that stochastic gradient descent can approximate the Bayesian uncertainty in model parameters near local minima. In this work we develop a similar correspondence for minibatch natural gradient descent (NGD). We prove that for sufficiently small learning rates, if the model predictions on the training set approach the true conditional distribution of labels given inputs, the stationary distribution of minibatch NGD approaches a Bayesian posterior near local minima. The temperature $T = \epsilon N / (2B)$ is controlled by the learning rate $\epsilon$, training set size $N$ and batch size $B$. However minibatch NGD is not parameterisation invariant and it does not sample a valid posterior away from local minima. We therefore propose a novel optimiser, "stochastic NGD", which introduces the additional correction terms required to preserve both properties.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
The kernel of graph indices for vector search
Model-Based Real-Time Pose and Sag Estimation of Overhead Power Lines Using LiDAR for Drone Inspection
PocketVina Enables Scalable and Highly Accurate Physically Valid Docking through Multi-Pocket Conditioning
SWE-Factory: Your Automated Factory for Issue Resolution Training Data and Evaluation Benchmarks
Step-by-step Instructions and a Simple Tabular Output Format Improve the Dependency Parsing Accuracy of LLMs
Employing self-supervised learning models for cross-linguistic child speech maturity classification
토론
이 논문과 연결된 포럼 글이 아직 없습니다.