TinyTL - Reduce Activations, Not Trainable Parameters for Efficient On-Device Learning

---

Created: =dateformat(this.file.ctime,"dd MMM yyyy, hh:mm a") | Modified: =dateformat(this.file.mtime,"dd MMM yyyy, hh:mm a") Tags: knowledge,

follows:• We propose TinyTL, a novel transfer learning method to reduce the training memory footprint byan order of magnitude for efficient on-device learning . We systematically analyze the show annotation

vation).• We also introduce the lite residual module, a memory-efficient bias module to improve the modelcapacity with little memory overhead. • Extensive experiments on trans show annotation