Deep Unsupervised Learning for Generalized Assignment Problems: A Case-Study of User-Association in Wireless Networks.

摘要: There exists many resource allocation problems in the field of wireless communications which can be formulated as generalized assignment (GAP). GAP is a generic form linear sum problem (LSAP) and more challenging to solve owing presence both equality inequality constraints. We propose novel deep unsupervised learning (DUL) approach time-efficient manner. More specifically, we new that facilitates train neural network (DNN) using customized loss function. This function constitutes objective penalty terms corresponding Furthermore, employ Softmax activation at output DNN along with tensor splitting simplifies guarantees meet constraint. As case-study, consider typical user-association network, formulate it GAP, consequently our proposed DUL approach. Numerical results demonstrate provides near-optimal significantly lower time-complexity.