A Data-driven Market Simulator for Small Data Environments - SIAM FME Virtual Talk

Explore the cutting-edge intersection of deep learning and financial mathematics in this virtual talk from the SIAM Activity Group on Financial Mathematics and Engineering. Dive into the world of data-driven market simulation for small data environments, presented by Dr. Blanka Horvath from King's College London. Discover how Deep Hedging is revolutionizing financial market modeling and learn about the challenges and opportunities in neural network-based generative approaches for financial time series. Examine a parsimonious generative model that performs well with limited training data, and understand how rough paths theory combined with Variational Autoencoders can effectively encode and evaluate financial data. Gain insights into pricing and hedging considerations within a deep neural network framework and their connection to market generation. This comprehensive talk covers topics ranging from historical market models to advanced concepts like mixture models, signatures, and log signatures, concluding with discussions on data privacy and performance metrics.

Introduction
Blanca
My screen
Background
DataDriven Models
Mixture Models
Generative Models
Signatures
Variational Autoencoders
Performance Evaluation Metrics
Log Signatures
General Strategy
Numerical Results
Why is this necessary
Data Privacy
Questions
Performance metric
Closing