Learning Compressed Embeddings for On-Device Inference

no code implementations • 18 Mar 2022 • Niketan Pansare, Jay Katukuri, Aditya Arora, Frank Cipollone, Riyaaz Shaik, Noyan Tokgozoglu, Chandru Venkataraman

MEmCom outperforms the state-of-the-art techniques, which achieved 16%, 6%, 10%, and 8% relative loss in nDCG at the respective compression ratios.

Model Compression Recommendation Systems