Recurrent Neural Networks with more flexible memory: better predictions than rough volatility

TL;DR

This paper introduces extended recurrent neural networks with multiple timescales, significantly improving long-memory process predictions like asset volatility over traditional LSTMs, with faster training and better accuracy.

Contribution

The authors develop a flexible timescale extension to LSTMs, enhancing their ability to model long-memory processes and outperforming rough volatility models in prediction accuracy.

Findings

Extended LSTMs require half the training epochs of vanilla LSTMs.

Extended LSTMs show less variability in validation and test losses.

The best model outperforms rough volatility predictions by about 20%.

Abstract

We extend recurrent neural networks to include several flexible timescales for each dimension of their output, which mechanically improves their abilities to account for processes with long memory or with highly disparate time scales. We compare the ability of vanilla and extended long short term memory networks (LSTMs) to predict asset price volatility, known to have a long memory. Generally, the number of epochs needed to train extended LSTMs is divided by two, while the variation of validation and test losses among models with the same hyperparameters is much smaller. We also show that the model with the smallest validation loss systemically outperforms rough volatility predictions by about 20% when trained and tested on a dataset with multiple time series.