This repository includes my work drafts at Aalto University, Probabilistic Machine Learning group. The topic was Bayesian experimental design.
In specific, this work is on Mössbauer spectroscopy, which is a spectroscopic technique that is used to observe nuclear interactions.
Those interactions might be informative regarding the physical properties of a material.
Since the likelihood function was not present in the paper on which this work is based, it was not possible to use a standard Bayesian experimental design.
Instead, a likelihood-free Bayesian experimental design method based on a framework called "LFIRE" is used.
The main theoretical sources for this work were:
- C. Feng, ‘Optimal Bayesian experimental design in the presence of model error’, Massachusetts Institute of Technology, Center for Computational Engineering, 2015. (PDF)
- S. Kleinegesse and M. Gutmann, ‘Efficient Bayesian Experimental Design for Implicit Models’, 10 2018. (PDF)
- O. Thomas, R. Dutta, J. Corander, S. Kaski, and M. U. Gutmann, ‘Likelihood-free inference by ratio estimation’, 11 2016. (PDF)
Most of the code is based on S. Kleinegesse and M. Gutmann's work and especially on the code that is published by S. Kleinegesse:
You can check the sources folder to see all the papers used during the work.
The work is coded in Python. Jupyter notebooks were used as an experimentation tool.
In this work, Python libraries such as NumPy, matplotlib, SciPy, glmnet, GPyOpt were also used.