Archive for September, 2007
I just submitted a paper that includes some very strange results that I got when playing around with the different collaborative filtering techniques on the MovieLens dataset. The work was a direct follow up to the new similarity measure I wrote about in my previous post on privacy in distributed recommender systems, and begins by reshaping the way we think about (and try to visualise) collaborative filtering. (more…)
Claudio (from IIIA and sponsored by MyStrands) gave a very interesting talk about poolcasting (pdf of his slides). Poolcasting is a web radio in which individuals may join different channels. Those subscribed to a channel will listen to the same stream of songs. The problem is how to select the songs on that stream. Claudio did so by combining the preferences of a channel´s listeners using Case-Based Reasoning.
The same approach may be used for mobile music. A bar may decide to play songs depending on the preferences of its customers (and preferences may be elicited from the playlists that customers store on their mp3 players or mobile phones).
A couple of questions that might be of interest to some of us: what if listeners do not share many songs in their playlists? Would it be possible to factor listeners´ reputation (trust) in deciding which songs to play?
From here: “mobileCampLondon will be taking place the last weekend of September or Saturday the 29th and Sunday the 30th. We’ve got an eclectic mix of participants coming: from those working on the open source mobile platform OpenMoko and the team behind the Mobile Advocacy Toolkit to CuteCircuit – the designers behind the hug shirt and other wearable computing experiments”.
Sign up (for free) on the wiki. (Space limited to 100 people).
Here are some papers presented at Mobiquitous. I’ll introduce them by research areas.
At this time of year the return to school is heralded by the purchase of new school uniform for thousands of children. Trutex, a clothing supplier in the UK, has recently announced that it is considering including GPS tracking devices in future ranges of its uniform products following an on-line survey of parents.
Mobile phone GPS tracking services targeted at parents for monitoring the whereabouts of their children are inexpensive and becoming more widely available in the UK. Advertisements often exploit parental concerns, heightened by high profile incidents of child abduction and violence from street gangs. ARCH, Action on Rights for CHildren, has raised key privacy issues with the UK government including the fact that there is no statutory regulation covering these devices beyond the Data Protection Act 1998.
Ethical issues abound, can a child or young person really give consent freely to be monitored in such a situation? Inevitably, cases of misuse are starting to emerge. Last month a man in the U.S was accused of tracking a child by implanting a GPS device in a picture frame the boy owned.
Returning to the subject of “smart” school uniform, human nature may rule the idea impractical in the end. In my experience by week 2 of term many children have lost the most expensive items of clothing you bought them or are just walking around in someone else’s.
Recently I have been working with Dimitris Moustakas to create a set of animations from bluetooth network traces. The animations are based on a dataset managed by and downloaded from CRAWDAD. The data was originally collected as part of the Reality Mining Project at MIT in which 100 participants were given bluetooth-enabled mobile phones and encouraged to carry them around over the course of the 2004-2005 academic year. Special software on the phones recorded bluetooth connections between devices. Specifically, if a mobile phone was within range of another and a connection was established then the start and end times of the connection and the device identification number were logged. The visualisations were created using SoNIA, a Java-based social network animation tool.
Situation: Using mobile devices, such as smart phones, people may create and distribute different types of digital content (e.g., photos, videos). One of the problems is that digital content, being easy to create and replicate, may likely swamp users rather than informing them. To avoid that, users may run trust models on their mobile devices. A trust model is a piece of software that keeps track of who provides quality content and who does not.
Problem: Devices should be able to set their initial trust for other devices. One way of doing so is for devices to learn from their own past experiences. To see how, consider the following quotes about human trust: “We may initially trust or not trust those involved on our projects based on past experience”, and “If your boyfriend is unfaithful, you won’t initially trust the next man you date” Algorithms that model human trust on pervasive devices, one might say, ought to do the same thing – they should assign their initial trust upon `similar’ past experiences.
Existing Solutions: Existing solutions usually require an ontology upon which they decide which past experiences are similar, and, in so doing, they require both that the same ontology is shared by all users (which is hardly the case in reality) and that users agree on that ontology for good (ie, the ontology is not supposed to change over time)
Proposal: TRULLO gathers ratings of past experiences in a matrix, learns staticial “features” from that matrix, and combines those features to set initial trust values. It works quite well in a simulated antique market and its implementation is reasonably fast on a Nokia mobile phone.
Future: TRULLO does not work if one does not have past experiences. That is why we will propose a distributed trust propadation algorithm (pdf).