Summary In view of the problems of low routing efficiency, complex control process, and difficult network management in big data environment in the traditional integrated space‐terrestrial network, in the paper, we propose a satellite network architecture called software‐defined information centric satellite networking (SDICSN) based on software‐defined networking (SDN) and information‐centric networking (ICN), and we design a virtual node matrix routing algorithm (VNMR) under the SDICSN architecture. The SDICSN architecture realizes the flexibility of network management and business deployment through the features of the separation of forwarding and controlling by the SDN architecture and improves the response speed of requests in the network by the centric of “content” as the ICN idea. According to the periodicity and predictability of the satellite network, the VNMR algorithm obtains the routing matrix through the relative orientation of the source and destination nodes, thus reducing the spatial complexity of the input matrix of the Dijkstra algorithm and then reducing the time complexity of the routing algorithm. For forwarding information base (FIB), the mechanism of combination of event driven and polling can be quickly updated in real time. Finally, the advantages of the SDICSN architecture in routing efficiency, request delay, and request aggregation are verified by simulation.

The influence of multiple longitudinal mode pulses on laser-induced damage at the exit surface of fused silica is investigated under simultaneous exposure to multi-wavelength lasers. The typical damage morphologies are systematically observed after a series of experiments have been conducted on fused silica samples by the equipment for multi-wavelength damage test. By means of the calculation for incubation and expansion fluences determined from the diameter of the ring pattern, the role of the each wavelength laser in damage process has been identified. The results provide insight into the information on the precursor reactiveness under simultaneous exposure to dual-wavelength pulses.

Abstract:Multi-task learning can extract the correlation of multiple related machine learning problems to improve performance. This paper considers applying the multi-task learning method to learn a single task. We propose a new learning approach, which employs the mixture of expert model to divide a learning task into several related subtasks, and then uses the trace norm regularization to extract common feature representation of these sub-tasks. A nonlinear extension of this approach by using kernel is also provided. Experiments conducted on both simulated and real data sets demonstrate the advantage of the proposed approach.

No abstract