I’ve found this website and it’s really nice. Cooking for Engineers regularly presents recepies and cooking tips and present them in a pleasing and concise way. It offers a handy conversion utility, based on Google Calculator, and a nice little community based around “analytical food” (before the forums were hacked by a jerk). It seems that the updates are too slow, but the guy is just doing this as a hobby, so he’s forgiven… 
June 27th, 2005 | Personal | No comments
Recently, a study was conducted by Hungarian scientist Z. Dezs? and his team on trying to quantitize visitation patterns in a rapidly changing network. Previous sutdies mostly focused on systems that stay relatively stable in time. However, a big part of the web is not made like this. The most common type of rapidly changing systems on the web are news portals and e-commerce sites. Understanding the dynamics of information access can yield tremendous insight to web designers and anyone working in the information technology field.
The main site that was studied is popular Hungarian news portal origo.hu. On this page, the stories are added and removed rapidly and gets roughly 6,5 E 6 hits per day. The team used the logfiles of the site to track the visitors individually on the course of one day and to track the number of documents released during this period (3908).
Fistly, they analysed the topology of the network by comparing the typical form of the graph of hits vs. time for a “skeleton” node and a “news” node. They then tried to build a model that represent the access of a single news document over the course of one day. hey have found that the majority of visits (28%) take place within the first day, rapidly decaying the next day but keeping a low but steady traffic generated by direct searches in the archives or traffic from outside links.
With a simple visitation pattern that they have concieved, they managed to represent pretty accurately the access of one document over time. As they analysed the data coming from the website, they have found that the hits pattern was of the power-law form and not the expected exponential decay.
Their results show that visitation of a document significantly decays after a 36 hour period. Moreover, a typical user sees only 53% of all news items appareaing on the main page (as they are replaced by newer ones, the news are moved to more specific pages, then are archived) and reads only 7% of them. Such shallow penetration was expected, but never quantified that precisely.
These decay laws are not only limited to news portals. They are likely generic because they only depend on the visitaiton pattern of a typical user. It could be applied to e-commerce, but the items are usually sold before we can take any conclusion from the law. The results might be applicable to biological systems too, to model the interactions of the molecules inside a cell.
This paper is currently in pre-print and can be accessed at http://xxx.lanl.gov/abs/physics/0505087
June 16th, 2005 | Earth Sciences and Geomatics, Physics | No comments