[Adapt] [Seminar]Model compression
gxq961127 at sjtu.edu.cn
Tue Oct 8 19:22:21 CST 2019
In deep learning, we may face this problem -- If we choose a large model the response time may be too long, but if we choose a small model the accuracy is not satisfactory. This is why we need model compression. Model compression aims to compress a complex large model into a simple small model with competitive performance. In this seminar, I will talk about model compression and go through a relevant paper: Rocket Launching: A Universal and Efficient Framework for Training Well-performing Light Net (Zhou et al. 2018, AAAI).
Time: Wed 4:30pm
Venue: SEIEE 3-414
See you there!
More information about the Adapt