Archive for January, 2007

Learning and Inferring Transportation Routines

Wednesday, January 31st, 2007

This paper (whose extended journal version has been recently accepted) proposes a hierarchical Markov model that learns and infers a user’s daily movements.

Portable Reputations

Friday, January 19th, 2007

In UTIFORO (a new research project), we may explore how sellers may “port” their reputation from eBay to informal markets. That might relate to this:
Last year I mentioned eBay’s Feedback system and said it was arguably their biggest asset. Even with its flaws, I said, it is one the biggest drivers of trust between two people buying and selling who’ve never met and never will. But it’s a closed system, usable only within eBay and only for eBay transactions.
We needed an internet-wide identity and feedback system that any reputable application can tap into, both pulling and pushing data.
At the time we had taken a look at iKarma, but they seemed to have missed the boat by ignoring the portability aspect of reputation.
Rapleaf launched in April. And while it’s still quite early, it does exactly what we need it to do – provide a good off-ebay reputation system. eBay banned Rapleaf in May (They learned their lesson with PayPal it seems), but the company is still chugging along.

Nissan tests intelligent transportation system

Friday, January 19th, 2007

by Jonathan Goddard

Nissan tests intelligent transportation system Image
Nissan tests intelligent transportation system Image Nissan tests intelligent transportation system Image

See larger / more images / slideshow

20 September 2006 – Nissan is testing a newly developed intelligent transportation system, which it is hoped will allow vehicle-to-infrastructure communication to reduce traffic accidents and ease congestion.

The test, slated to begin on 1 October and conclude in 2009, is in Kanagaw – 25km from Tokyo. Around 10,000 drivers, who must be subscribers to Nissan’s CARWINGS navigation service, are expected to participate.

The advanced road traffic system uses information obtained from nearby vehicles and roadside optical beacons to alert drivers to potential danger from approaching vehicles. The information is received by an onboard antenna on the vehicle, and the system uses the information to warn drivers when they are speeding in restricted zones. It also provides drivers with fastest-route information using “probe data” – information on the position and speed of vehicles obtained by wireless communications technology.

Based on the results of the test, Nissan is planning to implement its intelligent transportation system in Japan and then globally in the future as part of its efforts to help reduce traffic accidents and congestion. In Japan, Nissan has set a target of halving the number of traffic fatalities or serious injuries involving Nissan vehicles by 2015 compared with the 1995 level

Modelling the Spread of Rumours in Complex Social Networks

Thursday, January 18th, 2007

A while back Mazir gave an interesting talk whose abstract is below. One of the directions of future research is to model the spread of rumours in a web of trust. That might be of interest to some of us.

Rumours are an important form of social communications, and their
spreading plays a significant role in a variety of human affairs. Standard
models of rumours do not take into account the topology of the
underlying social interaction networks along which rumours spread. In
this talk I introduce a new stochastic model for the spread of rumours
in large-scale online social networks which are characterized by highly
complex connectivity patterns. I derive a set of mean-field equations
for the dynamics of the model on such networks and use analytical and
numerical solutions of these equations to examine the critical
properties and dynamics of rumour spreading on several models of
complex social networks.

Bluetooth and GPS on chip

Tuesday, January 16th, 2007

There is an interesting press release covered at VNUnet, claiming CSR will have an IC performing both Bluetooth and GPS out by the end of the year. Good news for power consumption and deployment issues.

Trust and reputation at AAMAS

Monday, January 15th, 2007

The program of AAMAS is out. Some of us may be interested in:

An Incentive Mechanism for Message Relaying in Unstructured Peer-to-Peer Systems
Information Searching and Sharing in Large-Scale Dynamic Networks
A Multi-Agent System for Building Dynamic Ontologies
Incentive Compatible Ranking Systems
On the Benefits of Cheating by Self-Interested Agents in Vehicular Networks
An Agent-Based Approach for Privacy-Preserving Recommender Systems
Effective Tag Mechanisms for Evolving Coordination
Distributed Task Allocation in Social Networks
Selecting Trust Evidences: an intuitive approach based on presumptive reasoning and domain analysis
Dynamically Learning Sources of Trust Information: Experience vs. Reputation
Rumours and Reputation: Evaluating Multi-Dimensional Trust within a Decentralised Reputation System

Animal Tags for People?

Friday, January 12th, 2007

Two cousin companies bet the fast-expanding market for animal RFID chips will extend to humans before long
… Digital tags are expected to be affixed to the U.S.’s 40 million farm animals to enable regulators to track and respond quickly to disease, bioterrorism, and other calamities. Opponents have many fears about this plan, among them that it could be the forerunner of a similar system for humans. …

Well, all you conspiracy buffs, let me introduce you to Kevin McGrath and Scott Silverman. … more

Scale-free networks emerging from weighted random graphs

Friday, January 5th, 2007

An alternative explanation for the emergence of scale-free degree distributions: http://polymer.bu.edu/hes/articles/ksbbhs06.pdf

A uniformly random weight is assigned to each edge in a classical random graph. Nodes connected by edges with weights less than the graph’s percolation threshold are collapsed into supernodes. The resulting graph has a power law degree distribution.

BitTyrant: a selfish BitTorrent client that improves performance

Friday, January 5th, 2007

BitTyrant is a BitTorrent client with a novel unchoking algorithm.

Suppose your upload capacity is 50 KBps. If you’ve unchoked 5 peers, existing clients will send each peer 10 KBps, independent of the rate each is sending to you. In contrast, BitTyrant will rank all peers by their receive / sent ratios, preferentially unchoking those peers with high ratios.

During evaluation testing on more than 100 real BitTorrent swarms, BitTyrant provided an average 70% download performance increase when compared to the existing Azureus 2.5 implementation, with some downloads finishing more than three times as quickly.

I wonder how well it performs in swarms of other BitTyrant clients?

The USENIX paper is here.

Update: it seems to be using the same faster than the bear algorithm I came up with last year. Damn it, I should have tried it out. :-)