Unverified50% confidenceSolutionExact time
加法节点的梯度传播规则是L对任一加数的偏导为1，直接将全局梯度传下去；乘法节点的偏导等于另一因子的值再乘以全局梯度
1
Sources
50%
Confidence
Long-term
Relevance
8/13/2026
First Seen
Sources
Related Claims
VerifiedThe essence of backpropagation is chain-rule differentiation, where the gradient of the loss function with respect to the parameters of each layer is propagated layer by layer from the output layer to the input layer, on the premise that the loss function must exist and be differentiable.65% similarUnverifiedREINFORCE估计器的方差极高，因为一条轨迹中的每个动作都乘以整条轨迹的总回报59% similarVerifiedWhen loss.backward() is called, PyTorch traverses the computation graph in reverse, using the chain rule to automatically calculate the gradient of each parameter.59% similarUnverifiedThe efficiency of backpropagation comes from reusing intermediate values and error terms stored during the forward pass—the core idea of dynamic programming—allowing gradients for all parameters to be obtained in a single backward pass.59% similarUnverifiedBackpropagation is essentially a special case of reverse-mode automatic differentiation; for a function with n input parameters and 1 scalar output, all partial derivatives can be computed with just one backward pass.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/741886API
curl https://kongchang.com/api/v1/knowledge/claims/741886MCP
get_claim(id=741886)