This paper presents a general half quadratic framework for simultaneous analysis of the whole array com- parative genomic hybridization (aCGH) profiles in a data set. The proposed framework accommodates different M-estimation loss functions and two underlying assumptions for aCGH profiles of a data set: sparsity and low rank. Using M-estimation loss functions, this framework is more robust to various types of noise and outliers. The solution of the proposed framework is given by half quadratic (HQ) mini- mization. To hasten this procedure, accelerated proximal gradient (APG) is utilized. Experimental results support the robustness of the proposed framework in comparison to the state-of-the-art algorithms.

Keywords