Dynamic L1-Norm Tucker Tensor Decomposition

Tucker decomposition is a standard method for processing multi-way (tensor) measurements and finds many applications in machine learning data mining, among other fields. When tensor arrive streaming fashion or are too to jointly decompose, incremental analysis preferred. In addition, dynamic adaptation of bases desired when the nominal subspaces change. At same time, it has been documented that outliers can significantly compromise performance existing methods analysis. this work, we present Dynamic L1-Tucker: an algorithm outlier-resistant data. Our experimental studies on both real synthetic datasets corroborate proposed (i) attains high estimation performance, (ii) identifies/rejects outliers, (iii) adapts changes subspaces.