Re-parameterizing Your Optimizers rather than Architectures

The well-designed structures in neural networks reflect the prior knowledge incorporated into the models. However, though different models have various priors, we are used to training them with model-agnostic optimizers such as SGD. In this paper, we propose to incorporate model-specific prior knowledge into optimizers by modifying the gradients according to a set of model-specific hyper-parameters. Such a methodology is referred to as Gradient Re-parameterization, and the optimizers are named RepOptimizers. For the extreme simplicity of model structure, we focus on a VGG-style plain model and showcase that such a simple model trained with a RepOptimizer, which is referred to as RepOpt-VGG, performs on par with or better than the recent well-designed models. From a practical perspective, RepOpt-VGG is a favorable base model because of its simple structure, high inference speed and training efficiency. Compared to Structural Re-parameterization, which adds priors into models via constructing extra training-time structures, RepOptimizers require no extra forward/backward computations and solve the problem of quantization. We hope to spark further research beyond the realms of model structure design. Code and models https://github.com/DingXiaoH/RepOptimizers . * Equal contributions. This work was partly done during their internships at MEGVII Technology. † Project leader. ‡ Corresponding author. 1 Prior knowledge refers to all information about the problem and the training data Krupka & Tishby (2007). Since we have not encountered any data sample while designing the model, the structural designs can be regarded as some inductive biases Mitchell (1980), which reflect our prior knowledge.