he-y / soft-filter-pruning

Soft Filter Pruning for Accelerating Deep Convolutional Neural Networks
https://arxiv.org/abs/1808.06866
376 stars 74 forks source link

How to deal with the situation where there is not enough zero filters? #8

Closed lkyf closed 5 years ago

lkyf commented 5 years ago

If we cannot obtain N{i+1}(1-P{i}) zero filters at layer i in the converged network. How to deal with it?

he-y commented 5 years ago

Assume we have 64 filters in the network. If the pruning rate is 30%, we select 19 filters with small L2-norm and prune them, in other words, we change all the value of those 19 filters to zero. If we set a threshold, we could always find them.