**The algorithm**

In our previous academic map, we mainly use Gephi to produce our maps. It has some limitations: Gephi can not handle big data layout problem(up to 0.4M), so we must separate our data into several parts according to their fields and visualize them individually. But there are lots of papers that belong to several fields, so it will cause ambiguity. At the same time, we can not represent the relation among fields using the layout of the fields.

The generation of the layout of the new graph is done using an N-body simulation similar to that used to simulate galaxy formation in astrophysics. Computing is naively an N^2 operation, and this calculation dominates each iteration of the simulation. In order to make it run at reasonable speed the Barnes-Hut algorithm is used. At each iteration a quad-tree is built out of all the nodes (taking order N log N time) and then this quad-tree is used to compute an approximation (taking order N log N time again).

**How to use**

The top layer show the relation between field. For instance, you will find Mathematics and Computer Science share some similarities and highly influence each other.

The bottom layer show the relation between several papers that you want. The color of link means when the citation relation was established. And the redius of paper means the influence of the paper(the more citation the paper has, the larger the radius will be).

If you want to search the information of a particular paper, just click on it. Then you will see some extra information of the paper including the title, the author, citation count, reference count and so on.

You can also search where the related papers are by clicking the citation link or reference link.

See more details in Academic Map. If you have any question or advice, please contact Wjerry5@sjtu.edu.cn

]]>**(1) Conference Author's Comparison Map**

In the author's comparison map, we selected the top 1,000 authors from each conference, and attach them with a red dot and a green dot. The yellow dots represent the authors that published papers in both conferences. If two authors have published a paper together in two conferences, the points would be linked by an edge. The more papers that have been published together, the greater the weight of the edge, the closer the two points.

Fig.1:SIGMOD & VLDB Author Comparison Map

Fig.2:SIGMOD & SIGIR Author Comparison Map

SIGMOD and VLDB are two type A conferences of the category of "Database/Data Mining/Content Search" in CCF conference category. The Fig.1 is SIGMOD&VLDB author comparison map. It is easy to figure from the map that yellow dots occupy the dominance, Which means most of top authors in both conferences also published papers in another conference. Additionally, The range and types of papers that SIGMOD and VLDB archived resembles and the authors between two conferences have firm contact.

SIGMOD and SIGIR are also two type A conferences of the category of "Database/Data Mining/Content Search" in CCF conference category. However, from the Fig.2, we could see that the red dots and great dots are separated conspicuously. Only few of the yellow dots in the boundary of two conferences. The reason of this phenomenon is that although SIGMOD and SIGIR are classified to the same category. The type of papers and range are different. SIGMOD conference inclines to Database while SIGIR inclines to Information Search.

Additionally, in the author comparison paper, you could use the size and color of dots to compare the influence of two conferences. In the 1,000 selected authors, if authors in a conference have more influence, their dots are decorated with red while others are decorated with green. The size of the dots means the total citations of the papers published by authors in two conferences.

For example, in Fig.3, INFOCOM & MOBICOM author comparison map, the red dots represent INFOCOM and the green dots represent MOBICOM. So we could draw the conclusion that the 1000 authors in INFOCOM have much more influence than the 1000 authors in MOBICOM. However, we could see that considering the size of the dots, green dots has more big size dots than red dots which means that some authors only published papers in MOBICOM have huge influence but INFOCOM did not have such authors.

Fig.3: INFOCOM & MOBICOM Author Comparison Map

**(2) Conference Papers' Comparison Map**

In conference papers' comparison map, we selected 500 papers from each conference which have high citations and use red and green dots to illustrated them. The size of the dots means the citation. Within the 500 papers, if a conference's total citations are larger than another one, it could be decorated with red. If a paper was cited by another paper in the map, the two papers would be linked by an edge.

Fig.4: MICRO & PLOS

Fig.5: POPL & SOSP

Fig.6: POPL & ASE

From the conference papers' comparison map, we could discover some phenomenons:

(1) Overlap Ornament: For example, in MICRO & ASPLOS conference papers' comparison map(Fig.4), the two conferences belong to "Computer Architecture/Parallel and Distributed Computation/Storage System" in CCF conference category. In this map, there is no obvious boundary between two different colored dots that they are firmly mixed which means that the papers between two conference have strong relationship. Also, it means that types and range of two conferences are similar. The total amount of red dots which represent MICRO are larger than ASPLOS (green dots). So it means the papers in MICRO have much more influence than papers in ASPLOS.

(2) One Dominance: POPL&SOSP papers' comparison map(Fig.5) belongs to "Software Engineering/System Software/Programming Language". The center of the map are dominated by red dots and red dots are larger than green dots. Little green dots are distributed in the outside circle. So it is easy to know that the POPL dominated and is more influential than SOSP.

(3) Hemispheric separation: POPL&ASE(Fig.6) belongs to "Software Engineering/System Software/Programming Language". The papers' comparison map has a boundary between two clusters, like the Hemisphere of earth. It shows that the citations between two conferences are not large, the contact between two conferences is not strong.

]]>We have analyzed and visualized the connections among paper in each two different academic sub-fields under

Please visit http://acemap.sjtu.edu.cn/app/AcademicMap/ and http://acemap.sjtu.edu.cn/typical for more details.

]]>Mentorship relationship is a special relationship between academic cooperation. To observe the mentorship of top authors of certain areas, we draw the maps of Mentorship Map (http://acemap.sjtu.edu.cn/topic/topicpage?topicID=0271BC14).

In studying the relationship between the figure, the author can observe the top authors’ mentorship and study the relationship between them, for instance, to view whom their mentors were, to observe their students’ academic achievement and whether their students are growing into a new generation of mentors and flourishing the mentorship network.

Mentorship relationship is based on the concentration, disparity, dominant degree of specificity, and other characteristics in cooperation, gained through machine learning prediction.

In studying the relationship in the figure, the nodes represent the authors, the size indicates the number of publication, the links represent the mentorship and the color represents the author in algebra from the map.

**2. Mentorship Hierarchical Map**

To observe the mentorship in the dimension of time, we draw the Mentorship Hierarchical Map.

Mentorship Hierarchical Map is based on the two-dimensional layout of Mentorship Map, added with the time axis. The top are the earliest scholars, and the bottom are the new scholars.

In studying the relationship between the layered graph, in addition to the scholars’ tutors and students, we can also observe the longitudinal comparison, in which historical period the top author's academic career started, which period gave birth to the authors with the most fruitful achievements, the distribution of top authors’ tutors and students, what time the students achieved more publication; can also make horizontal comparison, to compare the authors' academic achievements and the number of students of the same time.

In studying the relationship in the figure, the nodes represent the authors, the size indicates the number of publication, the links represent the mentorship and the color represents the author in what period of the history.

]]>This list adopts the algorithm designed by Acemap team to find high influential researchers of each university based on the dataset of Acemap, containing over 127M papers from over 24K venues. And ChinaSo team provides some related technical supports.

]]>This ranking is based on PaperRank, a paper ranking method based on random walk just like PageRank. First we normalize the rank of papers of each affiliation to 1-100(smaller means better). Then we calculate the score of each researcher of this affiliation by summing the reciprocals of his/her papers published in this affiliation. Finally we get the rank of researchers by sorting their scores descending.

]]>You can get information of researchers in the left panel by clicking corresponding nodes, jump to other researcher's Co-author map by double clicking corresponding nodes, and highlight some researchers according their affiliations.

There are more interesting functions, just try it!

]]>