Deep Neural Networks (DNNs) have revolutionized numerous aspects of daily life, yet their design and training processes remain inherently complex and often unpredictable. To address these challenges and provide systematic guidance for the development of DNNs, hyperparameter optimization (HPO) has emerged as a critical research focus. This paper investigates prominent methodologies for hyperparameter optimization, beginning with an analysis of essential hyperparameters and their pivotal influence on neural network performance. We then evaluate state-of-the-art hyperparameter optimization algorithms, assessing their computational efficiency, and highlighting their scalability issues. To overcome these issues, we discuss and study key distributed HPO strategies using the Ray framework. Through extensive experimentation, we rigorously evaluate these algorithms across a spectrum of hyperparameter configurations, demonstrating their scalability and effectiveness in optimizing model performance.