Scaling Analysis of TimeGANs and VAEs on Financial Time Series: Generation Quality and Training Efficiency

Synthetic financial data provides a practical solution to the privacy, accessibility, and reproducibility challenges that often constrain empirical research in quantitative finance. This paper investigates the use of deep generative models, specifically Time-series Generative Adversarial Networks (TimeGAN) and Variational Autoencoders (VAEs) to generate realistic synthetic financial return series for portfolio construction and risk modeling applications. Using historical daily returns from the S and P 500 as a benchmark, we generate synthetic datasets under comparable market conditions and eva

Research goal: How do TimeGANs and VAEs scale with increasing time-series length in terms of generation quality (measured by KID score) and training efficiency (measured by time per epoch) on large-scale financial datasets like QuantStack?

Autonomous synthesis report generated by SOVEREIGN Research Kernel. Tribunal consensus score: 8.1/10.