Smartformer: An Intelligent Transformer Compression Framework for Time-Series Modeling

Abstract

Transformer, as one of the cutting-edge deep neural networks (DNNs), has achieved outstanding performance in time-series data analysis. However, this model usually requires massive amounts of parameters to fit. Over-parameterization not only brings storage challenges in a resource-limited setting but also inevitably results in the model over-fitting. Even though literature works introduced several ways to reduce the parameter size of Transformers, none of them addressed this over-parameterized issue by concurrently achieving the following three objectives: preserving the model architecture, maintaining the model performance, and reducing the model complexity (number of parameters). In this study, we propose an intelligent model compression framework, Smartformer, by incorporating reinforcement learning and CP-decomposition techniques to satisfy the aforementioned three objectives. In the experiment, we apply Smartformer and five baseline methods to two existing time-series Transformer models for model compression. The results demonstrate that our proposed Smartformer is the only method that consistently generates the compressed model on various scenarios by satisfying the three objectives. In particular, the Smartformer can mitigate the overfitting issue and thus improve the accuracy of the existing time-series models in all scenarios.