基于Python分析学生学习算法研究与实践

Analysis of the Student Learning Algorithms in Python

基于Python语言,结合“老师-学生”模型以及卷积自编码网络提出了Sound Auto Encoder算法,深入研究学生学习算法和实践研究,通过无监督特征学习的方式处理音频数据,根据自编码网络、卷积神经网络原理及相关工作,提出SoundAutoEncoder模型及SoundNet中的“老师-学生”模型.在音频数据的特征学习中,通过分析现有算法优势及局限,对SoundAutoEncoder算法模型网络结构及网络学习算法进行分析.通过实验,对算法Sound Auto Encoder和Sound Net进行对比,在五折上,Soimd Auto Encoder算法取得的结果要比Sound Net明显好;在低于600次迭代中,Soimd Auto Encoder算法的结果相较好,Sound Net算法基本处于训练初级阶段,且与可获得的最好结果尚有差距.

A Sound Auto Encoder algorithm in Python is proposed in this paper by combining convolutional self-coding network and "teacher-student" model on the basis of in-depth study of student learning algorithms and practice research, in which unsupervised feature learning for audio data is carried out. In feature learning of audio data, the network structure of Sound Auto Encoder algorithm model and network learning algorithm are analyzed in terms of the advantages and limitations of existing algorithms. Comparison between Sound Auto Encoder algorithm and Sound Net algorithm through experiments shows that the result of Soimd Auto Encoder algorithm is better than that of Sound Net with a 50% discount;in less than 600 iterations, Soimd Auto Encoder algorithm achieves satisfactory result, but what Sound Net algorithm achieves is far from the best available results because it is basically in the initial stage of training.