Machine Learning for Pricing, Calibration and Hedging

• Deep Pricing
  • Consider the application of deep neural networks for pricing, e.g. applying the methods to enhance the speed of traditional VaR engines by replacing costly valuation functions with pre-trained neural networks
  • Proxy and missing data
• Deep Calibration
  • Applying deep neural networks for the calibration of financial models
  • Extend the scope of the models to those that were to costly to evaluate because the only available methods are simulation techniques
  • Optimize the neural network geometries for this task and provide stable, fast applications that could be used in the financial industry in production
• Deep Hedging
  • Applying deep neural networks (LSTM and ResNets) for the risk mitigation in trading
  • Taking into account transaction or funding costs
  • Extending hedge strategies beyond standard Greeks
• Pricing
  • Interpolation with regard to given dependence structures and constraints on interpolators
  “Interpolating” missing data with regard to main characteristics of the available data Time Series analysis – prediction, extreme value theory, etc. Interpolating/Extrapolating missing data