Generating Financial Time Series by Matching Random Convolutional Features
2026-06-03 • Machine Learning
Machine Learning
AI summaryⓘ
The authors tackle the problem of creating realistic financial time series data when only one limited historical example is available, which often leads to overfitting. To improve this, they propose training models to match random convolutional features of real and generated sequences rather than traditional fixed features that may miss important details. They introduce SOCK, a new fully differentiable feature mapping method suited to training these generative models. Their results show that SOCK-trained models perform better than previous methods on small financial datasets and also work well for other tasks like classification and hypothesis testing.
financial time seriesgenerative modelsoverfittingrandom convolutional featuresSOCKpath signaturesdiscriminatorfeature mapstwo-sample hypothesis testingtime series classification
Authors
Konrad J. Mueller, Nikita Zozoulenko, Ben Wood, Thomas Cass, Lukas Gonon
Abstract
Generating realistic financial time series is challenging as training data is often limited to a single historical path. With such scarce data, overfitting is hard to avoid, especially under adversarial training where a trained discriminator can memorize the training samples. To mitigate this, recent approaches train generators to minimize the discrepancy between untrained feature representations of real and generated time series. In these works, the feature maps are based on path signatures, which can fail to capture relevant time series properties at tractable truncation depths. In this work, we instead train generators by matching random convolutional features of real and generated time series. Existing random convolutional feature maps, such as Rocket and Hydra, have been shown to provide informative representations of real-world time series, but cannot supervise generative models because they are non-differentiable. We introduce SOCK (SOft Competing Kernels), a fully differentiable random convolutional feature map, suited to train generative time series models. We show that generators trained by matching random SOCK features consistently outperform signature and diffusion baselines across a wide range of small-sample financial datasets. We further demonstrate SOCK's expressiveness on two-sample hypothesis testing and time series classification tasks, where SOCK matches or outperforms existing unsupervised feature maps.