提交 bd14660c 编写于 作者: D dongzhihong

"add part of design doc"

上级 f5636dab
## Backward Policy Design
## Operator/expression 's Backward
### Motivation
In Neural Network, the backpropagation algorithm follows the chain rule, so we need to compound the fundmental gradient operators/expressions together with chain rule . Every forward network need a backward network to construct the full computation lineage, the operator/ expression's Backward feature will generate the backward pass respect to forward pass.
### Implement : gradient operator registry
| | forward operator | backward operator |
| ---------------------- | ---------------- | -------------------------------- |
| **Operator::inputs_** | Inputs | Inputs, Outputs, OutputGradients |
| **Operator::outputs_** | Outputs | InputGradients |
Inputs/Outputs means the input/output of the operator, InputGradients/OutputGradients is the gradient respect to forward opeartor. Forward operator and Backward operator are isomorphic, save their corresponding needs into member attribute.
We use a global hash map record the gradient operators available, follow the philosophy of minimum core, make operator pluggable unit. Each gradient is an operator and it needs to regist itself.
grad_op_builder(fengjiayi)
### Implement : Backward network
given a forward network, it generates the backward network. We only care about the Gradients—`OutputGradients`,`InputGradients`.
1. bla bla bla (yuyang)
2. NetOp
when the input forward network is a NetOp, it need to call the sub NetOp/Operators backward function recursively and ensure them done. During the process, we need to collect the `OutputGradients` name.
We share variable in the same scope, as a result, duplicate operator `OutputGradients` will overwirte then duplicate variable.
![./images/duplicate_op]()
Share variable between operators or same input variable used in multiple operators lead to a duplicate gradient variable. As demo show above, we need to rename gradient name recursively, and add a generic add operator instead.
![./images/duplicate_op2]()
​ Then collect the sub graph OutputGradients/InputGradients as the NetOp's and return it.
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册