Unverified50% confidenceFactExact time
A randomly initialized neural network of width m converges to a Gaussian process as m tends to infinity.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn Bayesian neural networks with continuously increasing width M, the KL divergence regularization term in the ELBO gradually overwhelms the expected log-likelihood term, causing the variational predictive distribution to degenerate into the prior predictive distribution.73% similarUnverifiedIn recurrent neural networks, the critical state corresponds to the maximum singular value (spectral radius) of the Jacobian matrix approaching 1.66% similarUnverifiedStochastic gradient descent divides data into mini-batches of size m to approximate the gradient, enabling n/m updates per epoch instead of one.64% similarUnverified深度学习中的随机种子控制权重初始化、数据加载批次打乱顺序、Dropout神经元随机失活以及优化器中的随机梯度估计等多个随机性来源63% similarUnverifiedAs network width M increases, the KL term grows linearly with the number of parameters while the expected log-likelihood term grows relatively more slowly, which is the fundamental mathematical cause of the prior dominance phenomenon.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1022953API
curl https://kongchang.com/api/v1/knowledge/claims/1022953MCP
get_claim(id=1022953)