Statistical Analysis Of NIPS Conference Papers Based On Document Feature Space

With the recent years of neural networks and machine learning have being hotly debated,how to analyze the development of its related areas of the status and research hot spots become particularly important,not only related to the machine learning and neural network of their own academic research,but also related to most of the application of science and technology in real life.Neural Information Processing Systems Conference(NIPS),as a top conference on machine learning and computational neuroscience,contains a large number of papers on machine learning and neural networks.This paper chooses the papers collected in the NIPS conference as the research object,the statistical analysis of the citation,word frequency and the subject based on the document feature space,we can get the development course,academic hotspot and subject direction of the neural network and machine learning field.The main work is as follows:Firstly,the NIPS conference papers related to the data collection.Crawl data on the NIPS official website of the paper through Python,extract the relevant author of the paper,title,summary,full text,etc.To crawl on Microsoft's academic NIPS conference papers through Python,extract the author of the paper,cited,citing,research field and so on.Secondly,citation analysis on the NIPS conference papers,there are qualitative and quantitative analysis from two perspectives of the author and the affiliation.According to the author's affiliation,the selected indicators include the affiliation?s articles,the amount of cited,cited per paper.According to the author,the selected indicators include the author's articles,the amount of cited,cited per paper,H indicators.It is expected that the NIPS conference papers will be analyzed from multiple perspectives,to know the main research directions in this area,and the distribution of the various affiliation and authors.Thirdly,the frequency analysis and subject analysis of NIPS conference papers are as follows.The word frequency analysis of this paper is based on the abstract of the paper.The abstract as the subject and the core content of the paper not only reduces the interference of some irrelevant words,but also reduces the amount of work on the degree;word frequency analysis is published on the annual papers?s abstract of the word frequency statistics,extracting the higher frequency words as the conference?s hot words.The subject analysis is based on the full text of the paper,using Latent Dirichlet Allocation(LDA)and Principal Component Analysis(PCA)related methods to analyze;which involves the LDA theme analysis is the use of the potential semantics of the paper that is different from the word frequency statistics,and it can semantically regulate the topic of the paper.Each paper can generate one or more topics,and the whole thesis is expressed by the relationship between them;The topic analysis on PCA is to use the principal component analysis method to reduce the dimensionality of the document feature matrix of NIPS Proceedings in order to understand the distribution of each keyword.Through citation analysis,word frequency analysis and topic analysis of NIPS conference papers,it can be found that the academic field includes psychology,computer vision,cognitive science,statistical linguistics and information theory in addition to machine learning and neuroscience.Algorithm optimization,bayesian method,kernel method and other theories are still the focus of scholars,these branches as the basis for machine learning and neural network applications,has been the researchers continue to improve and enhance.In the machine learning and neural network frontier development,deep learning has been widely studied and applied in language recognition,image recognition,and enhanced learning.