Path-Space Model Risk via Signature-Induced Optimal Transport

Battle Count: 129

Highly relevant for quantitative risk management in trading. The framework directly addresses model risk in path-dependent derivative pricing (Asian, lookback, barrier options), tail risk estimation, and stress testing. The budget-aware signature surrogate approach provides a practical workflow for robust pricing and risk assessment. The connection to exotic option pricing model risk (e.g., 40% price differences between local vol and jump models) is directly relevant to trading desk risk management. However, the framework is more suited for risk assessment and stress testing than for real-time trading signal generation.

High complexity due to multiple components: (1) computing geometric path signatures via iisignature library, (2) solving multi-budget optimal transport problems using Sinkhorn entropic regularization, (3) implementing budget-weighted lasso for sparse signature surrogate fitting, (4) calibrating coordinate-wise budgets through empirical OT, (5) evaluating shifted-threshold probability bounds via bisection, (6) handling MAD normalization of signature coordinates, (7) managing the three-stage workflow (budget calibration, score specification, robust evaluation). The paper provides detailed numerical parameters but no public code repository.

Code is available from the authors upon reasonable request. The paper provides detailed numerical setups including specific parameters, simulation settings (20,000 paths, 2,000 time steps, 5 replicates), and uses the open-source iisignature library. However, no public repository is linked. The theoretical proofs are complete in appendices. Budget calibration methodology is described but requires empirical OT solvers (Sinkhorn) that add complexity.