Stochastic Downsampling for Cost-Adjustable Inference and Improved Regularization in Convolutional Networks
This addresses the need for adaptable inference costs in CNNs for scenarios with variable computational budgets, offering a novel method for cost-adjustable inference and regularization.
The paper tackles the problem of training convolutional networks with adjustable computational costs for inference, proposing Stochastic Downsampling Point (SDPoint) to enable cost-adjustable inference and improve regularization, achieving competitive accuracy on image classification tasks.
It is desirable to train convolutional networks (CNNs) to run more efficiently during inference. In many cases however, the computational budget that the system has for inference cannot be known beforehand during training, or the inference budget is dependent on the changing real-time resource availability. Thus, it is inadequate to train just inference-efficient CNNs, whose inference costs are not adjustable and cannot adapt to varied inference budgets. We propose a novel approach for cost-adjustable inference in CNNs - Stochastic Downsampling Point (SDPoint). During training, SDPoint applies feature map downsampling to a random point in the layer hierarchy, with a random downsampling ratio. The different stochastic downsampling configurations known as SDPoint instances (of the same model) have computational costs different from each other, while being trained to minimize the same prediction loss. Sharing network parameters across different instances provides significant regularization boost. During inference, one may handpick a SDPoint instance that best fits the inference budget. The effectiveness of SDPoint, as both a cost-adjustable inference approach and a regularizer, is validated through extensive experiments on image classification.