Minimal implementation and experiments of "No-Transaction Band Network: A Neural Network Architecture for Efficient Deep Hedging".

Overview

No-Transaction Band Network:
A Neural Network Architecture for Efficient Deep Hedging

Open In Colab

Minimal implementation and experiments of "No-Transaction Band Network: A Neural Network Architecture for Efficient Deep Hedging".

Hedging and pricing financial derivatives while taking into account transaction costs is a tough task. Since the hedging optimization is computationally expensive or even inaccessible, risk premiums of derivatives are often overpriced. This problem prevents the liquid offering of financial derivatives.

Our proposal, "No-Transaction Band Network", enables precise hedging with much fewer simulations. This improvement leads to the offering of cheaper risk premiums and thus liquidizes the derivative market. We believe that our proposal brings the data-driven derivative business via "Deep Hedging" much closer to practical applications.

Summary

  • Deep Hedging is a deep learning-based framework to hedge financial derivatives.
  • However, a hedging strategy is hard to train due to the action dependence, i.e., an appropriate hedging action at the next step depends on the current action.
  • We propose a "No-Transaction Band Network" to overcome this issue.
  • This network circumvents the action-dependence and facilitates quick and precise hedging.

Motivation and Result

Hedging financial derivatives (exotic options in particular) in the presence of transaction cost is a hard task.

In the absence of transaction cost, the perfect hedge is accessible based on the Black-Scholes model. The real market, in contrast, always involves transaction cost and thereby makes hedging optimization much more challenging. Since the analytic formulas (such as the Black-Scholes formula of European option) are no longer available in such a market, human traders may hedge and then price derivatives based on their experiences.

Deep Hedging is a ground-breaking framework to automate and optimize such operations. In this framework, a neural network is trained to hedge derivatives so that it minimizes a proper risk measure. However, training in deep hedging suffers difficulty of action dependence since an appropriate action at the next step depends on the current action.

So, we propose "No-Transaction Band Network" for efficient deep hedging. This architecture circumvents the complication to facilitate quick training and better hedging.

loss_lookback

The learning histories above demonstrate that the no-transaction band network can be trained much quicker than the ordinary feed-forward network (See our paper for details).

price_lookback

The figure above plots the derivative price (technically derivative price spreads, which are prices subtracted by that without transaction cost) as a function of the transaction cost. The no-transaction-band network attains cheaper prices than the ordinary network and an approximate analytic formula.

Proposed Architecture: No-Transaction Band Network

The following figures show the schematic diagrams of the neural network which was originally proposed in Deep Hedging (left) and the no-transaction band network (right).

nn

  • The original network:
    • The input of the neural network uses the current hedge ratio (δ_ti) as well as other information (I_ti).
    • Since the input includes the current action δ_ti, this network suffers the complication of action-dependence.
  • The no-transaction band network:
    • This architecture computes "no-transaction band" [b_l, b_u] by a neural network and then gets the next hedge ratio by clamping the current hedge ratio inside this band.
    • Since the input of the neural network does not use the current action, this architecture can circumvent the action-dependence and facilitate training.

Give it a Try!

Open In Colab

You can try out the efficacy of No-Transaction Band Network on a Jupyter Notebook: main.ipynb.

As you can see there, the no-transaction-band can be implemented by simply adding one special layer to an arbitrary neural network.

A comprehensive library for Deep Hedging, pfhedge, is available on PyPI.

References

  • Shota Imaki, Kentaro Imajo, Katsuya Ito, Kentaro Minami and Kei Nakagawa, "No-Transaction Band Network: A Neural Network Architecture for Efficient Deep Hedging". arXiv:2103.01775 [q-fin.CP].
  • 今木翔太, 今城健太郎, 伊藤克哉, 南賢太郎, 中川慧, "効率的な Deep Hedging のためのニューラルネットワーク構造", 人工知能学 金融情報学研究会(SIG-FIN)第 26 回研究会.
  • Hans Bühler, Lukas Gonon, Josef Teichmann and Ben Wood, "Deep hedging". Quantitative Finance, 2019, 19, 1271–1291. arXiv:1609.05213 [q-fin.CP].
StackNet is a computational, scalable and analytical Meta modelling framework

StackNet This repository contains StackNet Meta modelling methodology (and software) which is part of my work as a PhD Student in the computer science

Marios Michailidis 1.3k Dec 15, 2022
Modified prey-predator system - Modified prey–predator model describes the rate of change for each species by adding coupling terms.

Modified prey-predator system We aim to study the behaviors of the modified prey–predator model and establish the effects of several parameters that p

Seoyoung Oh 1 Jan 02, 2022
"Neural Turing Machine" in Tensorflow

Neural Turing Machine in Tensorflow Tensorflow implementation of Neural Turing Machine. This implementation uses an LSTM controller. NTM models with m

Taehoon Kim 1k Dec 06, 2022
Tools for manipulating UVs in the Blender viewport.

UV Tool Suite for Blender A set of tools to make editing UVs easier in Blender. These tools can be accessed wither through the Kitfox - UV panel on th

35 Oct 29, 2022
Simple transformer model for CIFAR10

CIFAR-Transformer Simple transformer model for CIFAR10. Reference: https://www.tensorflow.org/text/tutorials/transformer https://github.com/huggingfac

9 Nov 07, 2022
Angora is a mutation-based fuzzer. The main goal of Angora is to increase branch coverage by solving path constraints without symbolic execution.

Angora Angora is a mutation-based coverage guided fuzzer. The main goal of Angora is to increase branch coverage by solving path constraints without s

833 Jan 07, 2023
Refactoring dalle-pytorch and taming-transformers for TPU VM

Text-to-Image Translation (DALL-E) for TPU in Pytorch Refactoring Taming Transformers and DALLE-pytorch for TPU VM with Pytorch Lightning Requirements

Kim, Taehoon 61 Nov 07, 2022
Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more

JAX: Autograd and XLA Quickstart | Transformations | Install guide | Neural net libraries | Change logs | Reference docs | Code search News: JAX tops

Google 21.3k Jan 01, 2023
Self-Supervised Multi-Frame Monocular Scene Flow (CVPR 2021)

Self-Supervised Multi-Frame Monocular Scene Flow 3D visualization of estimated depth and scene flow (overlayed with input image) from temporally conse

Visual Inference Lab @TU Darmstadt 85 Dec 22, 2022
Shitty gaze mouse controller

demo.mp4 shitty_gaze_mouse_cotroller install tensofflow, cv2 run the main.py and as it starts it will collect data so first raise your left eyebrow(bo

16 Aug 30, 2022
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

Super Resolution Examples We run this script under TensorFlow 2.0 and the TensorLayer2.0+. For TensorLayer 1.4 version, please check release. 🚀 🚀 🚀

TensorLayer Community 2.9k Jan 08, 2023
This code is an implementation for Singing TTS.

MLP Singer This code is an implementation for Singing TTS. The algorithm is based on the following papers: Tae, J., Kim, H., & Lee, Y. (2021). MLP Sin

Heejo You 22 Dec 23, 2022
Jupyter notebooks showing best practices for using cx_Oracle, the Python DB API for Oracle Database

Python cx_Oracle Notebooks, 2022 The repository contains Jupyter notebooks showing best practices for using cx_Oracle, the Python DB API for Oracle Da

Christopher Jones 13 Dec 15, 2022
A PyTorch Lightning solution to training OpenAI's CLIP from scratch.

train-CLIP 📎 A PyTorch Lightning solution to training CLIP from scratch. Goal ⚽ Our aim is to create an easy to use Lightning implementation of OpenA

Cade Gordon 396 Dec 30, 2022
MinkLoc3D-SI: 3D LiDAR place recognition with sparse convolutions,spherical coordinates, and intensity

MinkLoc3D-SI: 3D LiDAR place recognition with sparse convolutions,spherical coordinates, and intensity Introduction The 3D LiDAR place recognition aim

16 Dec 08, 2022
Code for CMaskTrack R-CNN (proposed in Occluded Video Instance Segmentation)

CMaskTrack R-CNN for OVIS This repo serves as the official code release of the CMaskTrack R-CNN model on the Occluded Video Instance Segmentation data

Q . J . Y 61 Nov 25, 2022
A PyTorch implementation of "Graph Wavelet Neural Network" (ICLR 2019)

Graph Wavelet Neural Network ⠀⠀ A PyTorch implementation of Graph Wavelet Neural Network (ICLR 2019). Abstract We present graph wavelet neural network

Benedek Rozemberczki 490 Dec 16, 2022
PyTorch implementation of a Real-ESRGAN model trained on custom dataset

Real-ESRGAN PyTorch implementation of a Real-ESRGAN model trained on custom dataset. This model shows better results on faces compared to the original

Sber AI 160 Jan 04, 2023
Official PyTorch implementation of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image", ICCV 2019

PoseNet of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image" Introduction This repo is official Py

Gyeongsik Moon 677 Dec 25, 2022
Multi-query Video Retreival

Multi-query Video Retreival

Princeton Visual AI Lab 17 Nov 22, 2022