Papers by Author: Liang Lei

Paper TitlePage

Authors: Liang Lei, Jun Peng
Abstract: The evaluation on the students' vocational skills and teaching quality is more often unilaterally performed by the school side. The results are not recognized by business, industry and other employers, and further to affect the students' employment rate and quality. This work team carried out a series of research subjects on how to establish a mechanism for third-party evaluation of teaching, where businesses were invited to join the teaching management, with their views on graduates being collected for the initial establishment of an effective evaluation mechanism. This played a very important role in exercising and improving the overall quality of students.
5116
Authors: Bo Yang, Ning Li, Liang Lei, Xue Wang
Abstract: Three-layer BP neural network, particularly using Levenberg-Marquardt back-propagation with early stopping algorithm, is widely used in curve fitting, attributing to its fast speed and free from over-fitting. Hence, the trained network by Levenberg-Marquardt back-propagation was used for curve fitting of the radiation spectrum of blast furnace raceway. The results showed that Levenberg-Marquardt back-propagation with early stopping algorithm presented a better fitting ability. Additionally, the results of spectral fitting model showed that the blast furnace raceway had an effective radiation spectrum in the wavelength range from 420nm to 880nm, where the raceway could be considered as the gray body radiation.
197
Authors: Bo Yang, Ning Li, Xue Wang, Liang Lei
Abstract: Neural network with Levenberg-Marquardt back-propagation training is widely used in curve fitting, according to its fast speed and free from over-fitting. In order to solve the issue on local minimum that may be found in Levenberg-Marquardt back-propagation with early stopping, and to get optimum number of hidden neurons, the least mean test errors algorithm was used in repeatedly training the three-layer feed-forward network with variable structure. Furthermore, the trained network was used for curve fitting of the radiation spectrum of blast furnace raceway. The results on spectra modeling of blast furnace raceway showed that this algorithm presented a better fitting ability, characterized by the reservation of the details of the original spectra and better generalization ability.
245
Showing 1 to 3 of 3 Paper Titles