Skip to content

Latest commit

 

History

History
28 lines (16 loc) · 988 Bytes

README.md

File metadata and controls

28 lines (16 loc) · 988 Bytes

Introduction

The repository contain the source code for the research project Applying machin learning for troubleshooting credit exposure and XVA profiles conducted in 2018 and 2019.

At this moment, it only contain the Convolutional Neural Network work,

Model training

The model is trained using the GPU platform offered by https://www.floydhub.com/.

Training code job can be found in \jobs

Data used in the training are also uploaded to FloydHub,

One can use the below command to start the model training(it took approx four hours for a standard GPU offered by FloyHub)

floyd run --gpu --env tensorflow-1.11 --data cufezhusy/datasets/gengduoshuju/4:gengduoshuju 'python model_training.py'

Also if you have enough local computational power, can also try to run it locally.

The training results can be found in \train_result

Working paper

Working paper about this project is avaliable in SSRN https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3404863