
For FullText PDF, please login, if you are a member of IEICE,
or go to Pay Per View on menu list, if you are a nonmember of IEICE.

MaxMinDegree Neural Network for CentralizedDecentralized Collaborative Computing
Yiqiang SHENG Jinlin WANG Chaopeng LI Weining QI
Publication
IEICE TRANSACTIONS on Communications
Vol.E99B
No.4
pp.841848 Publication Date: 2016/04/01 Online ISSN: 17451345
DOI: 10.1587/transcom.2015ADP0013 Type of Manuscript: Special Section PAPER (Special Section on Autonomous Decentralized Systems Technologies and Applications for NextGeneration Social Infrastructure) Category: Keyword: big data, cloud computing, decentralized computing, collaborative computing, learning systems,
Full Text: PDF>>
Summary:
In this paper, we propose an undirected model of learning systems, named maxmindegree neural network, to realize centralizeddecentralized collaborative computing. The basic idea of the proposal is a maxmindegree constraint which extends a kdegree constraint to improve the communication cost, where k is a userdefined degree of neurons. The maxmindegree constraint is defined such that the degree of each neuron lies between k_{min} and k_{max}. Accordingly, the Boltzmann machine is a special case of the proposal with k_{min}=k_{max}=n, where n is the fullconnected degree of neurons. Evaluations show that the proposal is much better than a stateoftheart model of deep learning systems with respect to the communication cost. The cost of the above improvement is slower convergent speed with respect to data size, but it does not matter in the case of big data processing.

