Dynamic Deep Neural Networks: Optimizing Accuracy-Efficiency Trade-offs by Selective Execution
This work addresses computational bottlenecks for deploying deep learning models in resource-constrained environments, representing a novel method rather than an incremental improvement.
The paper tackles the problem of computational inefficiency in deep neural networks by introducing Dynamic Deep Neural Networks (D2NN), which selectively execute only a subset of neurons per input to prune unnecessary computation, resulting in improved accuracy-efficiency trade-offs as demonstrated through extensive experiments on image classification tasks.
We introduce Dynamic Deep Neural Networks (D2NN), a new type of feed-forward deep neural network that allows selective execution. Given an input, only a subset of D2NN neurons are executed, and the particular subset is determined by the D2NN itself. By pruning unnecessary computation depending on input, D2NNs provide a way to improve computational efficiency. To achieve dynamic selective execution, a D2NN 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 D2NN is trained end to end. Both regular and controller modules in a D2NN 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 D2NN architectures on image classification tasks, we demonstrate that D2NNs are general and flexible, and can effectively optimize accuracy-efficiency trade-offs.