Model-Compression/Lossless_Compression
We propose a lossless compression algorithm based on the NTK matrix for DNN. The compressed network yields asymptotically the same NTK as the original (dense and unquantized) network, with its weights and activations taking values only in {0, 1, -1} up to scaling.
PythonApache-2.0
Stargazers
- 0seba
- Aimtrueflw
- btma48
- bwnjnOEIM78 Nebula
- duyongqiHuazhong University of Science and Technology
- FeiLiuEM
- gly-no
- godfatherqyx
- huiliutong
- jaemyungkim
- JeffCarpenterCanada
- jinhaohust
- k4u5h1k
- Leeon-KBUAA
- Lmy0217MUSIC Lab@SZU
- nankingjingHUST
- npusidnuputit
- OedoSoldier
- pprpData Science and Analytic Thrust, Information Hub, HKUST(GZ)
- synctextDelft University of Technology
- wmker
- wsx000