Eigentaste: A Constant Time Collaborative Filtering Algorithm

Eigentaste is a collaborative filtering algorithm that uses universal queries to elicit real-valued user ratings on a common set of items and applies principal component analysis (PCA) to the resulting dense subset of the ratings matrix. PCA facilitates dimensionality reduction for offline clustering of users and rapid computation of recommendations. For a database of n users, standard nearest-neighbor techniques require O(n) processing time to compute recommendations, whereas Eigentaste requires O(1) (constant) time. We compare Eigentaste to alternative algorithms using data from Jester, an online joke recommending system.Jester has collected approximately 2,500,000 ratings from 57,000 users. We use the Normalized Mean Absolute Error (NMAE) measure to compare performance of different algorithms. In the Appendix we use Uniform and Normal distribution models to derive analytic estimates of NMAE when predictions are random. On the Jester dataset, Eigentaste computes recommendations two orders of magnitude faster with no loss of accuracy. Jester is online at: http://eigentaste.berkeley.edu     

Utility of the Spelling Sensitivity Score to analyze spellings of children with specific language impairment

The purpose of this study was to examine the utility of the Spelling Sensitivity Score (SSS) beyond percentage correct scoring in analyzing the spellings of children with specific language impairment (SLI). Participants were 31 children with SLI and 28 children with typical language in grades 2–4. Spellings of individual words were scored using two methods: (1) percentage correct and (2) SSS. Children with SLI scored lower than children with typical language when spelling was analyzed with percentage correct scoring and with SSS scoring. In addition, SSS scoring highlighted group differences in the nature of spelling errors. Children with SLI were more likely than children with typical language to omit elements and to represent elements with an illegal grapheme in words, whereas children with typical language were more likely than children with SLI to represent all elements with correct letters.     

Quantitative understanding of the fly casting stroke through measurements and robotic casting

A stumbling point for many learning to fly fish is being able to fly cast well enough to ‘present’ a fly to a feeding fish. Even among more skilled fly fishers, learning to fly cast better is often a bottleneck towards advancement in the sport. Fly casting instruction today relies on the visual inspection of the fly casting stroke and the resulting ‘loop’ of fly line. These qualitative visual cues help fly casting instructors spot the strengths and weakness in a student’s casting stroke. We introduce two novel technologies for fly casting instruction that result in aquantitative understanding of the fly casting stroke. In particular, we describe a novel robot that is able to replicate standard overhead casts. This achievement is made possible by measuring the casting stroke using a small (MEMS) rate gyro to record the angular velocity of the fly rod in the casting plane. The robot is then programmed to replicate this angular velocity history. Potential uses of this robot include an objective means to assess fly rod, fly line and leader designs, as well as new methods for fly casting instruction.