Dynamic Deep Neural Networks: Optimizing Accuracy-Efficiency Trade-Offs by Selective Execution

We introduce Dynamic Deep Neural Networks (D2 NN), a new type of feed-forward deep neural network that allows selective execution. Given an input, only a subset of D2 NN neurons are executed, and the particular subset is determined by the D2 NN itself. By pruning unnecessary computation depending on input, D2 NNs provide a way to improve computational efficiency. To achieve dynamic selective execution, a D2 NN augments a feed-forward deep neural network (directed acyclic graph of differentiable modules) with controller modules. Each controller module is a sub-network whose output is a decision that controls whether other modules can execute. A D2 NN is trained end to end. Both regular and controller modules in a D2 NN are learnable and are jointly trained to optimize both accuracy and efficiency. Such training is achieved by integrating backpropagation with reinforcement learning. With extensive experiments of various D2 NN architectures on image classification tasks, we demonstrate that D2 NNs are general and flexible, and can effectively optimize accuracy-efficiency trade-offs.

Authors

• Lanlan Liu
• Jia Deng

Keywords

No keywords are indexed for this paper.