A survey on recent advances in design and application of k -WTA models

The k-winners-take-all (k-WTA) mechanism constitutes a fundamental computational primitive for competitive selection, sparsity enforcement, and decision making in neural and multi-agent systems. Over the past two decades, a rich body of k-WTA models has emerged across multiple research communities, featuring diverse formulations, dynamical properties, and implementation paradigms. This paper presents a comprehensive survey of recent advances in the design and application of k-WTA models from a unified perspective. We first review the fundamental definitions of the k-WTA operation. Centralised k-WTA models are then systematically summarised from circuit, signal processing, and neural network viewpoints. Subsequently, distributed k-WTA models are reviewed. A comparative overview of representative distributed k-WTA models is provided to highlight common design principles and emerging trends. We also discuss open challenges and future research directions for k-WTA.