Sig-wasserstein GANs for time series generation (2021-11-01T00:00:00.000000Z)

TL;DR

The SigWGAN is developed by combining continuous-time stochastic models with the newly proposed signature W1 metric, which allows turning computationally challenging GAN min-max problem into supervised learning while generating high fidelity samples.

Abstract

Synthetic data is an emerging technology that can significantly accelerate the development and deployment of AI machine learning pipelines. In this work, we develop high-fidelity time-series generators, the SigWGAN, by combining continuous-time stochastic models with the newly proposed signature W1 metric. The former are the Logsig-RNN models based on the stochastic differential equations, whereas the latter originates from the universal and principled mathematical features to characterize the measure induced by time series. SigWGAN allows turning computationally challenging GAN min-max problem into supervised learning while generating high fidelity samples. We validate the proposed model on both synthetic data generated by popular quantitative risk models and empirical financial data. Codes are available at https://github.com/SigCGANs/Sig-Wasserstein-GANs.git

Authors

Hao Ni

2 papers

Lukasz Szpruch

2 papers

Magnus Wiese

2 papers

Sig-wasserstein GANs for time series generation

TL;DR

Abstract

Authors

References26 items

Logsig-RNN: a novel network for robust and efficient skeleton-based action recognition

Robust Pricing and Hedging via Neural SDEs

Generating synthetic data in finance: opportunities, challenges and pitfalls

A Data-Driven Market Simulator for Small Data Environments

Conditional Sig-Wasserstein GANs for Time Series Generation

Sig-SDEs model for quantitative finance

Neural Controlled Differential Equations for Irregular Time Series

A Generative Adversarial Network Approach to Calibration of Local Stochastic Volatility Models

GANs May Have No Nash Equilibria

Deep Hedging: Learning to Simulate Equity Option Markets

Learning stochastic differential equations using RNN with log signature features

Quant GANs: deep generation of financial time series

On Gradient Descent Ascent for Nonconvex-Concave Minimax Problems

Generative adversarial networks for financial trading strategies fine-tuning and combination

Signature Moments to Characterize Laws of Stochastic Processes

Privacy and Synthetic Datasets

The Limit Points of (Optimistic) Gradient Descent in Min-Max Optimization

Are GANs Created Equal? A Large-Scale Study

Training GANs with Optimism

Cycles in adversarial regularized learning

The uniqueness of signature problem in the non-Markov setting

Characteristic functions of measures on geometric rough paths

Uniqueness for the signature of a path of bounded variation and the reduced path group

A Generalised Signature Method for Multivariate Time Series Feature Extraction

On ﬁnding local nash equilibria (and only local nash equilibria) in zero-sum games

GENERATIVE ADVERSARIAL NETS

Field of Study

Journal Information

Name

Venue Information

Name

Type

URL

Alternate Names