Neural Score-Driven Volatility and Risk Forecasting
The results show that the neural score-driven models are competitive with Gaussian and Student-$t$ GAS benchmarks that require a prespecified parametric innovation distribution, with particularly strong performance during the COVID-19 period.