Department: Computer Science & Engineering
Faculty Advisor(s): William Griswold
Award(s): Honorable Mention

Primary Student
Name: Shiri Dori
Email: sdori@ucsd.edu
Phone: 617-792-5067
Grad Year: 2013

People rapidly learn the capabilities of a new location, without observing every service and product. Instead they map a few observations to familiar clusters of capabilities, and assume the availability of other capabilities in the cluster. This poster presents a similar approach to computer-based discovery of routine location capabilities, applying singular value decomposition to predict unobserved capabilities based on a combination of a small body of local observations and a larger body of data that is not specific to the location.

For shopping purposes, an area within easy walking distance is a single location, but may contain many different shops and services, collectively offering its own combination of capabilities. Truncated singular value decomposition (T-SVD) maps the observations to combinations of features, rather than to a single cluster.

Simulations, using distributions derived from real world data, demonstrate the feasibility of this approach. We show that using tapered SVD reduces the rank sensitivity, and prevents over- or under-fitting the data. The robustness of the T-SVD technique was further tested by introducing some erroneous data; enhancing the algorithm to accept user feedback enabled overcoming this challenge. The T-SVD technique also extends to estimate whether a capability is available at a given time.

The capabilities inferred from T-SVD could be used for context-aware reminders, using proximity detection or geofencing. Another application would be to plan shopping optimally according to the user's needs, e.g. by finding a single location which will have most of the products you are looking for; finding the closest set of stores that you could go to that would cover all your shopping needs; or finding the location with the cheapest products. Taking capabilities one step further, one could automatically compare pricing and find coupons appropriate to the shopping list.

« Back to Posters or Search Results