Rocket Launching: A Universal and Efficient Framework for Training Well-performing Light Net

Explore this paper's citation graph

Summary

This work proposes a universal framework that exploits a booster net to help train the lightweight net for prediction and uses one technique called gradient block to improve the performance of light net and booster net further.

Type
article
Published
2017-08-01
Cited by
129
References
34
Access
Open access

Keywords

Booster (rocketry), Computer science, Net (polyhedron), Inference, Benchmark (surveying)

References

Cited by

Related papers