Ph.D. Dissertation talk: Efficient Deep Neural Networks

Bichen Wu's Ph.D. dissertation talk at UC Berkeley -- 05/08/2019.


The success of deep neural networks (DNNs) is attributed to three factors: increased computing capacity, more complex models, and more data. These factors, however, are not always available especially for edge applications such as autonomous driving, augmented reality, and internet-of-things. Training DNNs requires a large amount of data, which are difficult to obtain. Edge devices such as mobile phones have limited computing capacity, therefore, require specialized and efficient DNNs. However, due to the enormous design space and prohibitive training costs, designing efficient DNNs for different target devices is challenging. So the question is, with limited data, computing capacity, and model complexity, can we still successfully apply deep neural networks?



In this talk, I will introduce our research effort to address the above problems and improve the efficiency of deep neural networks at many different levels. Model efficiency: we designed neural networks for various computer vision tasks and achieved more than 10x faster speed and lower energy. Data efficiency: we developed an advanced tool that enables 6.2x faster annotation on LiDAR point cloud. We also leveraged domain adaptation to utilize simulated data, bypassing the need for real data. Hardware-efficiency: we co-designed neural networks and hardware accelerators and achieved 11.6x faster inference. Design efficiency: the process of finding the optimal neural network is time-consuming. Our automated neural architecture search algorithms discovered, using 421x lower computational cost, models with state-of-the-art accuracy and efficiency.