I would like to check the possibility of implementation of BO using C language.
The implementation target is MCU like Tricore or further.
Is this possible?
Moreover, is there anyone know where the C-implementation sources is present at github or other sites?
Finally, what would be the most difficult parts while I try to implement BO to MCU starting from matlab toolbox or python code (@github) ?
I know that BO uses Gaussian Process Regression in order to estimate surrogate function from experiment data. Is this the most difficult part for implementation of BO?
Thanks in advance.
Dear Youngil,
I highly recommend the Matlab coder based on my experience. This package allows one to seamlessly generate C code from Matlab as long as no prohibited functions are used. I have used this to deploy code to MCUs on several occasions.
https://www.mathworks.com/products/matlab-coder.html
Refer to the PyMc package:
https://docs.pymc.io/en/v3/Gaussian_Processes.html
https://github.com/gbosquechacon/statrethink_course_in_pymc3
Statistical rethinking (2nd ed) focuses on Bayesian inference and the week 10 lecture material covers the Gaussian Processes.
https://machinelearningmastery.com/what-is-bayesian-optimization/
https://github.com/gbosquechacon/statrethink_course_in_pymc3
https://docs.pymc.io/en/v3/Gaussian_Processes.html
https://www.mathworks.com/products/matlab-coder.html
Hope it helps
Yes it’s quite possible to implement on any embedded system with pure C code, however you have to chose proper mcu, most difficult part would be utilizing cores efficiently since concurrency is not implemented as we used to be in modern processors. That’s why picking up right mcu is critical.
Python code might be easier for development stage, however i highly recommend pure C when it comes to mission-critical/production environment due to customized/optimized compilers from providers of embedded systems