论文部分内容阅读
针对高校科研水平深度学习网络训练评价中存在评价特征同质化现象,造成评估结果精度不高的问题,提出稀疏交叉熵粗糙集双向受限制深度玻尔兹曼机(DDRBM-DNNS)高校科研能力评估方法.首先,考虑采用受限制玻尔兹曼机(RBM)和稀疏交叉熵惩罚参数对深度学习网络进行改进,实现深度学习网络特征训练同质化现象的削弱;同时,针对输入数据的预处理问题,考虑基于粗糙集的前置预处理方式实现,在维持数据输入信息完整前提下,实现输入样本数据的有效归约,进而实现样本处理量的简化,有利于深度学习网络收敛过程的提速;最后,利用所提算法对高校科研水平进行评价,实验数据显示,所提评价模型具备更高的评估精度和更快运算效率.
In view of the existence of the phenomenon of homogeneity of the evaluation features in the evaluation of network-based deep learning network at universities and colleges, which results in the low accuracy of the evaluation results, the research capability of dual-direction restricted depth Boltzmann machines (DDRBM-DNNS) Evaluation method.First, consider the use of restricted Boltzmann machine (RBM) and sparse cross-entropy penalization parameters to improve the deep learning network to achieve the weakening of the homogeneity of the deep learning network feature training; at the same time, for the pre-input data To deal with the problem, the preprocessing based on rough set is considered. With the premise of maintaining the integrity of the data input information, the effective reduction of the input sample data is realized, so as to simplify the sample processing volume and speed up the deep learning process of network convergence Finally, using the proposed algorithm to evaluate the scientific research level in colleges and universities, experimental data show that the proposed evaluation model has higher evaluation accuracy and faster computing efficiency.