Title :
A Neural Network Pruning Method Optimized with PSO Algorithm
Author :
Tu, Juanjuan ; Zhan, Yongzhao ; Han, Fei
Author_Institution :
Sch. of Comput. Sci. & Telecommun. Eng., Jiangsu Univ., Zhenjiang, China
Abstract :
A neural network (NN) pruning method optimized with particle swarm optimization (PSO) algorithm is proposed in this paper. Correlation merging algorithm is an important pruning method in NN structure design. Unlike general training method with back-propagation (BP), this paper uses PSO algorithm in the pruning process. The PSO is used to optimize the initial parameters of the NN, including the weights and biases etc. The experiment results show that the method in the paper above conventional one has greater improvement in both accuracy and velocity of convergence for NN.
Keywords :
backpropagation; convergence; neural nets; particle swarm optimisation; PSO algorithm; backpropagation; convergence velocity; correlation merging algorithm; neural network pruning method; particle swarm optimization algorithm; Algorithm design and analysis; Computational modeling; Computer science; Information processing; Merging; Neural networks; Neurons; Optimization methods; Particle swarm optimization; Training data; neural network; particle swarm optimization; pruning;
Conference_Titel :
Computer Modeling and Simulation, 2010. ICCMS '10. Second International Conference on
Conference_Location :
Sanya, Hainan
Print_ISBN :
978-1-4244-5642-0
Electronic_ISBN :
978-1-4244-5643-7
DOI :
10.1109/ICCMS.2010.424
