Research Interests

• Model compression & acceleration
• Geometry theory
• Riemannian optimization
• Decentralized optimization

My primary research achievements to date have been associated with model compression and acceleration, through quantization and pruning techniques. I am now interested in Riemannian optimization through the lens of geometry theory. I believe that Riemannian geometry can provide effective tools for solving many problems in machine learning and provide theoretical explanations for the black box of deep learning.