This paper presents a comprehensive evaluation of modern deep learning methodologies for portfolio optimization, integrating Graph Neural Networks (GNNs), Deep Reinforcement Learning (DRL), Transformers, and Autoencoders with traditional financial models. Using historical data from 2015-2023 across equities, ETFs, and bonds, the study assesses predictive power for covariance estimation, return forecasting, dynamic asset allocation, and dimensionality reduction. Hybrid approaches such as Transformer+GNN and Autoencoder+DRL are introduced to capture both relational and temporal market structures. Performance is evaluated through backtesting with metrics including volatility, cumulative return, maximum drawdown, annualized return, and Sharpe ratio across seven strategies, including Equal-Weighted, 60/40 allocation, and Mean-Variance Optimization (MVO).

Key findings reveal that hybrid models, particularly Transformer+GNN, achieve superior stability and risk control, yielding the lowest volatility and drawdown. However, MVO paired with well-calibrated inputs delivers the highest cumulative return and Sharpe ratio, highlighting the continued relevance of traditional methods. Standalone DRL underperforms due to limited structural awareness, while Autoencoders exhibit behavior similar to Equal-Weight strategies, emphasizing the need for dynamic policy learning. The study aligns with existing literature on relational modeling and feature compression in finance, demonstrating that combining deep learning with financial theory yields robust and adaptive portfolio strategies. The authors suggest exploring latent representations within traditional optimization frameworks to improve scalability and performance.