Hybrid deep additive neural networks
This work addresses the issue of parameter efficiency and performance in neural networks for data science applications, but it appears incremental as it builds on existing additive regression and neural network architectures.
The authors tackled the problem of traditional neural networks requiring many parameters and sometimes having unsatisfactory performance by introducing hybrid deep additive neural networks that incorporate additive regression ideas. The result was that these networks generally achieved better performance than traditional neural networks while using fewer parameters, as shown in simulation studies and a real-data application.
Traditional neural networks (multi-layer perceptrons) have become an important tool in data science due to their success across a wide range of tasks. However, their performance is sometimes unsatisfactory, and they often require a large number of parameters, primarily due to their reliance on the linear combination structure. Meanwhile, additive regression has been a popular alternative to linear regression in statistics. In this work, we introduce novel deep neural networks that incorporate the idea of additive regression. Our neural networks share architectural similarities with Kolmogorov-Arnold networks but are based on simpler yet flexible activation and basis functions. Additionally, we introduce several hybrid neural networks that combine this architecture with that of traditional neural networks. We derive their universal approximation properties and demonstrate their effectiveness through simulation studies and a real-data application. The numerical results indicate that our neural networks generally achieve better performance than traditional neural networks while using fewer parameters.