The, surprisingly somewhat belated, report on the Granular Computing Conference GrC’09

Last week I attended the IEEE International Conference on Granular Computing 2009 in Nanchang, China. I had done my preparations to report about it also during the conference, but, alas, I was not allowed to even read my own blog, let alone posting to it; see [footnote 1] below for some observations on non-accessible blogs. This being the situation, hereby then a slightly belated report.

Tengwan Ge in Nanchang, which was also depicted on the beautiful GrC'09 remembrance plate from Jiangxi University

Tengwang Ge in Nanchang, which was also depicted on the beautiful GrC'09 remembrance plate from Jiangxi University

Keynotes

T.Y. Lin of San José State University, USA—one of the initiators of “Granular Computing” as a, in his view, specialisation area of applied mathematics—spoke about the difference between keyword terms in Web searches and the concepts behind it and he proposed to solve the issues with category theory. On my question why specifically category theory and not Semantic Web languages and technologies, I unfortunately did not get a clear answer. Xingdong Wu of the University of Vermont talked about mining user patterns, aggregations, and user interest modelling with wildcards, where the use of wildcards provide much flexibility (and improvements) in specifying and mining the user patterns. Last, Xue-wen Chen of the University of Kansas first went through the usual introductory aspects of system biology, to proceed to the actual topics of gene and protein networks. The technologies used are hidden Markov models, Bayesian networks, and a K-GIDDI divide-and-conquer biclustering algorithm that touched up with gene ontology terms to, respectively: find 4 new genes (their functions) in the D. melanogaster (fruitfly), figure out a gene network (80% of what was known already in the literature, but now done by automation), and 95.42% correct function assignment of H. sapiens genes (verified against the 437 genes already annotated) of an overall amount of 618 function assignments. Not bad, not bad at all; his 2008/9 papers have the details.

Sessions

My paper on granular perspectives ([1] and summary here) was scheduled right at the start in the first of the parallel sessions, in the “Foundations of granular computing”. It was listened to with interest, and I received positive feedback–if people will use it is, time will tell. Hong Hu brought up the topics of dynamic similarity (e.g. the gradual changes from tadpole to frog) and how to deal with it, for which he proposed to use neural networks [2]. However, dealing with the standard ‘static’ similarity, i.e. comparing two objects at the same time, is already a widely researched area, and unsurprisingly, the last word has not been said yet about dynamic similarity either; in fact, perhaps they were the first ones. The same session had scheduled a paper with a preliminary, set-based, notion for a theory of granularity [3], which already looks ahead to using attributes of the objects (but that was not fully integrated in the theory yet), and gives “granule” as a lump of objects in a granular level an explicit place in the theory. Yinliang Zhao’s paper was about a way to granulate program code, and in particular of object-oriented programming languages with a restriction, thus far, to single-inheritance class hierarchies (the programming version of taxonomies) [4].

In the afternoon, I went to the ‘Japanese session’, where, among others, Toyota presented an application paper for visualising the topics of and navigation to the 7000 or so Japanese laws to acquaint Japanese citizens with it (this year Japan changed its judicial system that now has a citizen-judge, or ‘saiban-in’ system) [5]. If this is practically scalable to the Italian system? It should, in theory at least, but with its more than 70000 laws, it will require more levels of granularity than the four they have so as to obtain appropriate overviews. In addition, the relations of the (hierarchically ordered) key terms in the texts of the full collection of Japanese laws have the characteristic of a so-called “small world network”, which makes it very suitable for visualisation. My experience with the Italian bureaucracy and its rules for the strangest things gives me the impression that that may not be the case with the Italian laws (and Italian laws can benefit from an automated consistency check, but that is a separate topic), but it is not a trivial exercise to actually verify or refute this hunch. As a tidbit of fun information about the relatedness of the keywords extracted from the Japanese laws: “Nation” scored highest with 1020 links, which was followed by “Money” with 981 links, which the presenter found curious enough to emphasise.

There were three sessions on rough sets: applications, theory, and computing. The applications session had a paper on using a “knowledge quantity” for relative importance of attributes used to compute the rough sets and to apply that to Chinese text categorisation using the “document frequency thresholding” characteristic [6]: while common terms appear to be important for global performance, rare terms “are the most informative” to be able to discern (make distinguishable) those documents from others and are, from a rough set perspective, therefore influential because they have most effect on the equivalence structure. Li [7], on the other hand, improved on the “extenics” company evaluation method by using rough sets so that the amount of company indicators, such as “human capital” and “technological innovation ability”, could be reduced, hence a company’s evaluation method simplified. On the theory side, there was, among others, a paper on neighborhood systems with respect to rough sets where a new “and” operator is introduced and, as the authors claim, is “different from traditional rough set approximations” [8]. The remainder of the paper to back up this claim is rather dense, but T.Y. Lin summarised his students’ work as that the lower and upper approximations in VPRS are special cases of the interior and closure in topological space. Last, Chen, Li and coauthors sought to dynamically update the upper and lower approximations of a rough set to reflect the changes in the underlying information system over time, and they presented the theory, algorithm, and experimental validation in [9,10].

Start of the walk at Lushan mountain, after the flower garden

Start of the walk at Lushan mountain, after the flower garden

The House where Nobel Priye winner Pearl Buckly lived on the Lushan Mountain

The House where Nobel Prize winner (literature, 1938) Pearl Buck lived on the Lushan Mountain

Other

As social event, besides the conference dinner, we had a trip to the Lushan mountain, which is a UNESCO world heritage site. Although I had to skip the walking sessions, the scenery is really beautiful and temperature comfortable. The visit to its “many old buildings” appeared to be the missionaries outpost of about 100-150 years ago, including a not very protestant church that is (still/again?) used for weddings, and the home of Pearl Sydenstricker Buck, who had won the Nobel prize for Literature for her writings about life in China.

Each participant also received a beautiful present from the local organisation, Jiangxi University of Finance and Economics: a black ceramic plate with in gold-coloured imprints the name of the university, IEEE GrC 2009, and in the centre the famous building Tengwang Ge.

Travelling to China is a bit of a hassle with the visa, and knowing some Chinese (which I do not) will be useful for getting around and things done, but nevertheless I highly recommend people to visit the country, be it a conference or holiday: the people are friendly and very helpful, the food is delicious, and there are lots of things to see and do.


References

  1. C. Maria Keet. From granulation hierarchy to granular perspective. In: Proceedings of the 5th IEEE international conference on Granular Computing 2009 (GrC’09). 17-19 August, Nanchang, China. IEEE Computer Society, 306-311.
  2. Hong Hu and Zhongzhi Shi. Machine learning as granular computing. In: Proc. of GrC’09. IEEE Computer Society, 229-234.
  3. Hong Li. Granule, Granular Set and Granular System. In: Proc. of GrC’09. IEEE Computer Society, 340-345.
  4. Yinliang Zhao. A step toward code granulation space. In: Proc. of GrC’09. IEEE Computer Society, 799-804.
  5. Tetsuya Toyota and Hajime Nobuhara. Hierarchical structure analysis and visualisation of Japanese law networks based on morphological analysis and granular computing. In: Proc. of GrC’09. IEEE Computer Society, 539-543.
  6. Yan Xu and Wang Bin. Knowledge management based on rough set. In: Proc. of GrC’09. IEEE Computer Society, 654-657.
  7. Yuan-yuan Li and Jun Yun. A comprehensive evaluation method based on extenics and rough set. In: Proc. of GrC’09. IEEE Computer Society, 381-383.
  8. Xibei Yang, Xinzhe Li and Tsau Young Lin. First GrC model — Neighborhood Systems: the most general rough set models. In: Proc. of GrC’09. IEEE Computer Society, 691-695.
  9. Weili Zou, Tianrui Li, Hongmei Chen, Xiaolan Ji. Approaches for incrementally updating approximations based on set-valued information systems while attribute values’ coarsening and refining. In: Proc. of GrC’09. IEEE Computer Society, 824-829.
  10. Hongmei Chen, Tianrui Li, Weibin Liu. Research on the approach of dynamically maintenance of approximations in rough set theory while attribute values coarsening and refining. In: Proc. of GrC’09. IEEE Computer Society, 45-48.

Notes

[footnote 1] Regular readers may recollect that Cuba did not block my blog, and that I have written a post there during the Informatica 2009 conference about the VIP session. This made me curious as to what type of blogs are (not) accessible here in China. Some observations (pages checked on 16 and 17 Aug 2009):

  1. WordPress: I did a random check of a few other wordpress blogs with full names as well as xxx.wordpress.com and .org types, such as Duncan‘s and WP’s own blog with tips ‘n tricks, to ascertain if it was just my blog being “timed out”, but all those blogs were “timed out”, too. I could access WP’s startpage.
  2. Blogspot: I tried Ben‘s and FSP’s blogs, which had a “connection interrupted” message. The http://www.blogger.com had a quick “connection interrupted” message, idem http://www.blogspot.com.
  3. Typepad: the frontpage already “timed out”, idem specific typepad blogs.
  4. Other blogs that do not run through one of those blogging sites but have their own software running, such as those of Michael Nielsen, LogBlog, and Microbeworld are accessible, but not the asmblog of the American Society for Microbiology, such as Small things considered (“timed out”, although the ASM was accessible).
  5. Curiously, when I did a Google search on “blog filters china”, one of the top hits returned was the accessible Harvard blog called “internet and democracy blog“, but the first hit returned by the Google search was a news item at National Public Radio that Microsoft implements blog filters for China, which closes with the line “Microsoft’s blogging filter could be seen as taking American companies’ cooperation with censorship to a new level. Instead of merely blocking what Internet users can read, she says, Microsoft is now limiting what they can write.”.

So, to whoever developed the filtering algorithms: there is room for improvement of your work; unless the owners of the three above-mentioned blogging softwares do this blocking themselves preemptively already, which I hope is not the case. To whomever who wants to have their blog also reach the Chinese in China: for the time being, install your own blogging software.