Urban vertical farming has emerged as a transformative solution to address food security challenges in densely populated cities by enabling year-round crop production within controlled environments. However, conventional layout design approaches often rely on trial-and-error methods that fail to optimize the complex interplay between spatial configuration, environmental parameters, and resource allocation. Machine learning offers powerful data-driven methodologies to systematically optimize vertical farm layouts by analyzing multidimensional datasets encompassing light distribution, airflow patterns, temperature gradients, humidity levels, and crop-specific growth requirements. This article examines the application of supervised, unsupervised, and reinforcement learning algorithms in optimizing rack configurations, crop placement strategies, and environmental control systems to maximize yield per unit area while minimizing energy and water consumption. Key machine learning techniques including artificial neural networks, genetic algorithms, support vector machines, and deep learning models are evaluated for their capacity to predict optimal spatial arrangements based on historical farm performance data and real-time sensor inputs. The integration of machine learning with digital twin technologies and computational fluid dynamics simulations enables dynamic layout reconfiguration responsive to changing crop growth stages and market demands. Applications demonstrate significant improvements in space utilization efficiency, energy savings through optimized LED placement, water conservation via precision irrigation scheduling, and enhanced crop productivity. This review highlights the critical role of machine learning-driven layout optimization in advancing sustainable urban food production systems and identifies future research directions for scalable implementation in commercial vertical farming operations.