Relax DARTS: Relaxing the Constraints of Differentiable Architecture Search for Eye Movement Recognition

TL;DR

Relax DARTS introduces an improved neural architecture search method for eye movement recognition, enabling more precise and adaptable models that outperform existing techniques on multiple datasets.

Contribution

The paper proposes Relax DARTS, a novel NAS algorithm that relaxes constraints of DARTS, allowing independent architecture training and flexible input selection for better performance.

Findings

Achieves state-of-the-art recognition accuracy on four databases.

Demonstrates adaptability to other multi-feature temporal classification tasks.

Improves training efficiency and model precision over traditional DARTS.

Abstract

Eye movement biometrics is a secure and innovative identification method. Deep learning methods have shown good performance, but their network architecture relies on manual design and combined priori knowledge. To address these issues, we introduce automated network search (NAS) algorithms to the field of eye movement recognition and present Relax DARTS, which is an improvement of the Differentiable Architecture Search (DARTS) to realize more efficient network search and training. The key idea is to circumvent the issue of weight sharing by independently training the architecture parameters $\alpha$ to achieve a more precise target architecture. Moreover, the introduction of module input weights $\beta$ allows cells the flexibility to select inputs, to alleviate the overfitting phenomenon and improve the model performance. Results on four public databases demonstrate that the Relax DARTS achieves state-of-the-art recognition performance. Notably, Relax DARTS exhibits adaptability to other multi-feature temporal classification tasks.