A Generalization Bound of Deep Neural Networks for Dependent Data
arXiv:2310.05892v12 citationsh-index: 14
Originality Synthesis-oriented
AI Analysis
This addresses a limitation in theoretical ML for applications like epidemiology and finance, but it is incremental as it extends existing bounds to dependent data.
The paper tackles the problem of generalization bounds for deep neural networks under non-iid data, establishing a bound for feed-forward networks with non-stationary φ-mixing data.
Existing generalization bounds for deep neural networks require data to be independent and identically distributed (iid). This assumption may not hold in real-life applications such as evolutionary biology, infectious disease epidemiology, and stock price prediction. This work establishes a generalization bound of feed-forward neural networks for non-stationary $φ$-mixing data.