Abstract—In this paper, we propose an algorithm to estimate missing values of electroencephalogram (EEG) signals in tensor form. EEG signals is a type of brain data which is recorded from many electrodes placed on the scalp. EEGs are easily missing values if recording in the bad condition or transforming from raw data into more understandable structure. The method is proposed to solve the problem in tensor form and it is built on recent studies about tensor completion using decomposition method. First, we give a definition for tensor completion using the component of decomposing process. Second, the tensor completion is formulated as a convex optimization problem. Therefore, a first order optimization method, Nonlinear Conjugate Gradient, is applied to solve it. In this paper, we also show several numerical experiments that demonstrate the effective of the proposed method.

Index Terms—Tucker decomposition, tensor completion, electroencephalogram (EEG), optimization.

Thao Nguyen Thieu, Hyung-Jeong Yang, Tien Duong Vu are with Dept. of Electronics and Computer Engineering, Chonnam National University, Gwangju, South Korea (e-mail: hjyang@jnu.ac.kr).
Sun-Hee Kim is with Dept. of Brain and Cognitive Engineering, Korea University, Seoul, South Korea.

Cite: Thao Nguyen Thieu, Hyung-Jeong Yang, Tien Duong Vu, Sun-Hee Kim, "Tensor Completion Based on Tucker Factorization for Missing Value Imputation," International Journal of Bioscience, Biochemistry and Bioinformatics vol. 6, no. 2, pp. 34-40, 2016.