Outlier-robust estimation methods for unnromalized models [1]
γ-noise contrastive estimation (GNCE)
Power noise contrastive estimation (PNCE)
Python implementation of GNCE and PNCE: robust-est.zip
demo.py and demo_cv.py are demos with and without cross-validation for γ and β, respectively.
For comparison, this implementation includes noise contrastive estimation [2] as well.
[1] Hiroaki Sasaki and Takashi Takenouchi, "Outlier-robust parameter estimation for unnormalized statistical models ", under review.
[2] Michael U. Gutmann, and Aapo Hyvärinen, "Noise-contrastive estimation of unnormalized statistical models, with applications to natural image statistics", Journal of Machine Learning Research, vol.13, pp.307–361, 2012