Contributions

Application of machine learning methods based on LAZE priors in cancer data

Time: 2017-11-02 21:42:28

       The research focuses on how to establish a sparse Bayesian model to analyze the prostate cancer data by machine learning methods. Based on the classical sparse methods, the Bayesian model based on a LAZE(Laplace with an Atom at ZEro, abbreviated as LAZE) prior is firstly applied in a cancer detection problem. A LAZE prior is a mixed distribution composed of a Laplace distribution and a Dirac distribution, that is, a Laplace distribution with an atom at zero. Moreover, we use the MCMC method to solve the Bayesian model based on a LAZE prior. For the cancer detection problem, a LAZE prior is more rubust than a Laplace prior in the Bayesian model. Finally, the numerical experiments of the prostate cancer data show that the Bayesian model based on a LAZE prior is better than another classical sparse algorithms.



References:

1. Ying Lin and Qi Ye. Application of Machine Learning Methods Based on LAZE Priors to Cancer Data -- Take the Prostate Cancer Data Set for Example. Journal of South China Normal University, 50(4), 115-120, 2018 (Chinese), DOI: 10.6054/j.jscnun.2018089