Identity Mappings in ResNet: The Residual Block
Beginner
The problem, from first principles
Residual networks (He et al., 2016) add the block's input back to its output, so each block learns a residual on top of an identity path. The identity shortcut is what lets very deep networks train.
From theory to code
Implement residual_forward(x, f), which returns x + f(x).
Constraints
fmust return the same shape asx.
Hints
Hint 1
Call f once and add the result to x.
Keywords
research-papersarchitectureresnet
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.