High Performance Mobile System for Deep Learning
- Distributed Deep Neural Network (DNN) system on mobile devices;
- Accelerated Generative Adversarial Networks (GAN) on mobile devices.
Low Power Mobile Computing Systems
- Real-time power consumption analysis for mobile system and applications;
- Performance retaining real-time CPU/GPU scheduling for mobile graphic system;
- User motion based low-frequency GPS sampling and route reconstruction system for low power navigation on wearable device.
Organic Ligh Emitting(OLED) Related Display Technologies
- Mobile OLED-friendly recording and playback system for low power video streaming;
- Dynamic tone mapping for OLED display power saving based on video classification;
- Low-power circuit design of OLED pixel for fine-grained dynamic voltage scaling;
- Sensing & compensation circuit design of OLED panel for display enhancement of non-uniformity & degradation.
Mobile Graphic Rendering Technologies
- Low-power VR/AR system design;
- Distributed parallel rendering framework for local wireless mobile device cluster;
- Quality Retaining Display Rendering Workload Scaling based on User-Smartphone Distance.
- Fast generation and detection framework of adversarial attack examples;
- Residual learning network for restoration of adversarial attack examples;
- Machine learning based user finger operation modeling, and its application on mobile device authentication.
Education and Others
Education Oreinted Project
Digital Circuit, CMOS VLSI and other Related Works
- Circuit modeling & performance analysis for STT-RAM crossbar circuits;
- ircuit implementation & performance analysis for classic neural synapse circuits;
- Chip tape-out of a 4K STT-RAM chip set, a joint project with SMIC and Chinese Academy of Science Physics Institute.