Professional Documents
Culture Documents
This paper introduces a novel implementation of the genetic algorithm exploiting a multi-GPU cluster. The
proposed implementation employs an island-based genetic algorithm where every GPU evolves a single
island. The individuals are processed by CUDA warps, which enables the solution of large knapsack instances
and eliminates undesirable thread divergence. The MPI interface is used to exchange genetic material among
isolated islands and collect statistical data. The characteristics of the proposed GAs are investigated on a two-
node cluster composed of 14 Fermi GPUs and 4 six-core Intel Xeon processors. The overall GPU performance
of the proposed GA reaches 5.67 TFLOPS.
An improved quantum genetic algorithm (IQGA) is proposed in this paper, which codes the chromosome with
probability amplitudes represented by sine and cosine functions,
and uses an adaptive strategy of the rotation angle to update the population. Then the mutation operation is
considered in this improved quantum genetic algorithm (MIQGA). Rapid convergence and good global search
capability characterize the performance of MIQGA. While testing, a variance function is introduced to
estimate the stability of the algorithm. When solving 0-1 knapsack problem, greedy repair function is used to
repair unfeasible solutions . Experimental results show MIQGA
has better comprehensive performance than traditional genetic algorithm (GA), standard quantum genetic
algorithm (QGA) and IQGA, especially the superiority in terms of optimization quality
and population diversity.
With the accelerated proliferation wireless networks ranging from GPRS and EDGE to high speed
networks such as HSPDA and Mobile Wi-Fi, network selection by mobile nodes will benefit more from
knowledge of Network Capability of candidate networks. Network selection is important for handover in
heterogeneous wireless environment. User Profiles/Needs and Network Capability
will greatly influence the next logical step after network discovery, which is Network Selection. We examine
the Dynamic Network Selection paradigm that uses User Profiling/needs to rank networks for selection and
ignore networks with less capacity than required , using the Knapsack problem 0/1 Dynamic algorithm and the
Knapsack problem Optimization Algorithm.
Author No Presensi
Hao Qian, Qingyong Zhang, Deming Lei, Zixiao 1
Pan School of Automation