Apr 5, 2017

Dilated convolution

Main idea:
"...it allows for very large receptive fields while only growing the number of parameters logarithmically. [Zhou at al.] shows how you can use a similar idea in classification to achieve 3.6× parameter reduction with only 1 drop of accuracy." - excerpt from Reference: dilated-convolutions-and-kronecker-factorisation/

More references

No comments: