Swish: How Close Is It to ReLU?
Beginner
The problem, from first principles
Swish is meant to behave like ReLU where ReLU works well, while staying smooth elsewhere. This question measures how far apart the two curves are, which is the quantity the paper's sharpness parameter controls.
From theory to code
Implement swish_relu_gap(x, beta), returning the largest absolute difference between swish(x, beta) and relu(x) over the input values.
Constraints
- Returns a Python float.
Hints
Hint 1
Compute both functions over the array, take the absolute difference, then the maximum.
Keywords
research-papersactivationswish
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.