<p>Multi-view feature selection aims to enhance the modeling capability of machine learning models for complex problems by integrating information from multiple views and selecting the most crucial and representative features. Current multi-view feature selection methods focus on the distribution and information differences brought by different views, while overlooking the issue of noise in multi-view data. In this paper, we propose a noise-tolerant multi-view feature selection algorithm. Specifically, we integrate view weights, sample weights, and feature weights into a unified framework to construct the objective function for multi-view feature selection. Subsequently, we employ a novel alternating iterative optimization algorithm to solve the proposed objective function and perform multi-view feature selection. During this process, noise samples are assigned smaller weights, enhancing the algorithm’s noise tolerance. Finally, a series of experiments are conducted on classification and regression tasks. The proposed algorithm achieves superior performance compared to state-of-the-art algorithms, particularly on datasets containing noise.</p>