Unverified50% confidenceFactExact time
In automatic differentiation, when a variable affects the loss function through multiple computation paths, the gradients from each path must be summed (gradient accumulation).
1
Sources
50%
Confidence
Long-term
Relevance
9/9/2026
First Seen
Sources
Related Entities
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.68% similarUnverified梯度下降及其变体（Adam、AdaGrad等）本质上是数值优化算法67% similarUnverified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66% similarUnverified多级调用图能展示函数的上游调用者和下游被调用者，帮助评估改动的潜在影响范围65% similarVerifiedWhen loss.backward() is called, PyTorch traverses the computation graph in reverse, using the chain rule to automatically calculate the gradient of each parameter.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/885582API
curl https://kongchang.com/api/v1/knowledge/claims/885582MCP
get_claim(id=885582)