google/tf-quant-finance

What it solves

This library provides high-performance components for quantitative finance, leveraging TensorFlow's hardware acceleration and automatic differentiation to implement mathematical methods and pricing models efficiently.

How it works

The library is organized into three tiers of functionality:

1. Foundational methods: Core math tools including optimization, interpolation, root finders, and random number generation.
2. Mid-level methods: Advanced tools such as ODE and PDE solvers, the Ito process framework, and Diffusion Path Generators.
3. Pricing methods: Specific quantitative finance utilities, including pricing models (e.g., Local Vol, Stochastic Vol, Hull-White) and their calibration, as well as rate curve building and payoff descriptions.

Who it’s for

It is designed for quantitative analysts and developers who need to perform complex financial modeling and pricing using the computational power of TensorFlow.

Highlights

• Hardware acceleration support via TensorFlow.
• Automatic differentiation for model calibration and sensitivity analysis.
• Support for a wide range of models including Heston, SABR, and Hull-White.
• Comprehensive examples covering American Option pricing, Monte Carlo simulations, and swap curve fitting.