The virtues of computational learning noise in volatile environments

This is a project realised at Ecole Normale Supérieure by Charles Findling under the supervisions of Nicolas Chopin and Etienne Koechlin.

Link to the paper

Briefly, the paper investigates the virtues of computational learning noise in volatile environments and shows it provides adaptive features.

Summary of the code

This code provides all models used in the paper:

• The computational varying volatility model
• The computational constant volatility model
• The algorithmic varying volatility model
• The algorithmic constant volatility model
• The algorithmic noise model
• The reinforcement model
• The noise-free PROBE model
• The noisy PROBE model

The volatility models as well as the PROBE models have their main workers coded in C++.

General arhitecture of the code:

• lib_c - containes C files for the volatility models and PROBE models
• fit_functions - contains fit functions for all volatility models and RL model
• simulation_functions - contains simulation functions for all volatility models and RL model
• utils - contains util python functions

Code compilation

To compile the c++ libraries, you will need to install the boost c++ library, version 1.59 - https://www.boost.org/users/history/version_1_59_0.html

Once downloaded, open the compile_c.sh file. Modify it by adding your boost library path. Then launch ./lib_c/compile_c.sh.

Enquiries

If you have any questions on this code or related, please do not hesitate to contact me: charles.findling(at)gmail.com