Unverified50% confidenceFactExact time
Gradient descent updates parameters along the opposite direction of the loss function's gradient, and the parameter update formula is usually w ← w − η·∂L/∂w, where η is the learning rate.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Backpropagation Explained: Principles and Code for the Core Algorithm Behind Large Models
bilibili大模型阿琴10/16/2026
Related Entities
Related Claims
UnverifiedThe partial derivative of the MSE loss function with respect to parameter W is ∂L/∂W = (2/n) * Xᵀ(XW - y), and the partial derivative with respect to b is ∂L/∂b = (2/n) * Σ(y_pred - y).69% similarUnverifiedIn robust regression, replacing the Gaussian likelihood function with a t-distribution is equivalent to reducing the weight of outliers in the loss function, which improves the model's robustness to anomalous values.66% similarUnverifiedL1 regularization adds the sum of the absolute values of parameters as a penalty term in the loss function, and its geometric properties cause the optimal solution to tend toward the coordinate axes; L2 regularization adds the sum of squared parameters, penalizing large coefficients without driving them to zero.62% similarUnverified策略梯度定理表明期望回报对策略参数的梯度可表示为 ∇J(θ) = E[∇log π_θ(a|s) · A(s,a)]，其中A是优势函数62% similarUnverifiedIn the selective prediction framework, the utility function is typically written as U = α·P(correct) - β·P(wrong), where α and β represent the reward weight for correct answers and the penalty weight for wrong answers, respectively.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1043699API
curl https://kongchang.com/api/v1/knowledge/claims/1043699MCP
get_claim(id=1043699)