ABCpy, a machine learning library developed by teams at the University of Oxford, Warwick University, Harvard University, and others has used γ-ABC, a python package proposed by our researchers* in the Imperfect Information Learning Team in RIKEN AIP and others as a part of their research paper titled “γ-ABC: Outlier-Robust Approximate Bayesian Computation based on A Robust Divergence Estimator” **accepted at AISTATS-21 this year.
In this paper, Fujisawa et al.(2021) propose a robust Approximate Bayesian Computation (ABC) that can achieve high accuracy even under contaminated data with a lot of noise or outliers. ABC is an inference scheme for models to simulate complex phenomena and is used in a wide range of scientific research fields, e.g., cosmology, genetics, and biology.
The integration of this method will allow many researchers and data analysts around the world working in a variety of scientific fields to use this method without implementation and to perform highly reliable simulations quickly.
*Authors:
Masahiro Fujisawa (The University of Tokyo / RIKEN)*Primary Author
Takeshi Teshima (The University of Tokyo / RIKEN)
Issei Sato (The university of Tokyo/RIKEN)
Masashi Sugiyama (RIKEN / The University of Tokyo)
（As of January 29, 2021）
**In Proceedings of 24th International Conference on Artificial Intelligence and Statistics (AISTATS2021), pp.1783-1791, online, Apr.
13-15, 2021.
For more details, please see the following website.
http://proceedings.mlr.press/v130/fujisawa21a.html
-ABCpy
https://abcpy.readthedocs.io/en/latest/abcpy.html#abcpy.distances.GammaDivergence
https://arxiv.org/pdf/2104.03889.pdf
https://abcpy.readthedocs.io/en/latest/abcpy.html
https://arxiv.org/pdf/1711.04694.pdf
