Accelerated Exotic Option Pricing

This project demonstrates two models for accelerating exotic option pricing using NVIDIA GPUs:

1. Monte Carlo Method (monte-carlo-method.cu)

• Implementation: CUDA C/C++
• Overview: Achieves optimal performance for pricing options by simulating millions of paths on GPUs.
• Details:
  • Requires explicit memory management and boilerplate code.
  • Python GPU libraries like CuPy and RAPIDS simplify the process, enabling scalable and distributed computations with minimal performance loss.

---

2. Deep Derivative Method (deep-derivative-method.py)

• Implementation: Python with PyTorch
• Overview: Uses a deep neural network to approximate option pricing for significant speed improvements.
• Details:
  • Delivers up to 35x speed improvements in inference time while maintaining accuracy.
  • Facilitates efficient calculation of Greeks using a fully differentiable neural network.
  • Further optimization achievable via TensorRT for production-grade performance.

---

These models highlight the power of GPU acceleration in quantitative finance, offering a blend of computational efficiency and simplicity for research and production environments.