Flexible sample selection strategies for transfer learning in ranking

Ranking is a central component in information retrieval systems; as such, many machine learning methods for building rankers have been developed in recent years. An open problem is transfer learning, i.e. how labeled training data from one domain/market can be used to build rankers for another. We propose a flexible transfer learning strategy based on sample selection. Source domain training samples are selected if the functional relationship between features and labels do not deviate much from that of the target domain. This is achieved through a novel application of recent advances from density ratio estimation. The approach is flexible, scalable, and modular. It allows many existing supervised rankers to be adapted to the transfer learning setting. Results on two datasets (Yahoo’s Learning to Rank Challenge and Microsoft’s LETOR data) show that the proposed method gives robust improvements.     

Career Narratives of African American Female College Students: Insights for College Counselors

The authors collected narrative data from 11 African American female college students at a predominantly White institution using the Career Construction Interview (Savickas & Hartung, 2012 ), the Future Career Autobiography (Rehfuss, 2009 ), and a qualitative career experiences questionnaire. Using the constant comparison method (Strauss & Corbin, 1990 ), the authors uncovered 4 overarching themes based on the unique experiences of participants. The authors present insights for college counselors providing career development services.     

Effects of turf and cleat footwear on plantar load distributions in adolescent American football players during resisted pushing

Metatarsal and midfoot injuries are common in American football. Footwear design may influence injury rates by altering plantar foot loading patterns in these regions. The purpose of this study was to determine the effect of cleat design on in-shoe plantar foot loading during a football-specific, resisted pushing task. Twenty competitive football players (age 14.7 ± 1.8 years, height 1.72 ± 0.10 m, and mass 71.8 ± 26.9 kg) completed three trials of pushing a weighted sled at maximal effort in a standard shoe (CLEAT) and artificial turf-specific shoe (TURF), with flexible in-shoe force measuring insoles. Repeated measures ANOVAs identified mean differences in maximum force and relative load under all regions of the foot. Results showed higher forces in the CLEAT under the medial (p < 0.001) and lateral (p = 0.004) midfoot, central (p = 0.007) and lateral (p < 0.001) forefoot, and lesser toes (p = 0.01), but lower forces in the hallux (p = 0.02) compared to the TURF shoe. Additionally, relative loading was higher in the CLEAT under the medial (p < 0.001) and lateral (p = 0.002) midfoot and lateral (p < 0.001) forefoot, but lower in the medial forefoot (p = 0.006) and hallux (p < 0.001) compared to the TURF shoe. The two shoes elicited distinct plantar loading profiles and may influence shoe selection decisions during injury prevention or rehabilitation practices.     

Reliability of a 5-m multiple shuttle test

The aim of the present study was to determine the reliability of a modified 5-m multiple shuttle test. The 'match-related fitness' of 23 female hockey players was assessed on four occasions within 4 weeks. The results of each test session and each shuttle were analysed using analysis of variance with repeated measures to determine the reliability of the test. The mean distance for each of the six shuttles decreased (121.2 +/- 7.5, 114.5 +/- 7.5, 112.2 +/- 7.5, 109.9 +/- 7.9, 108.4 +/- 8.1 and 108.7 +/- 8.3 m for shuttles 1-6, respectively; P < 0.001) similarly for each of the four sessions (P = 0.99). The total and peak distances covered during the tests were not significantly different (P = 0.99 and P = 0.12, respectively). The intra-class correlation coefficient (R) for these variables was 0.98 and 0.86, respectively. The delta distance and the fatigue index calculated post-test were significantly different (P = 0.001 and P = 0.006, respectively) between the four sessions. The intra-class correlation coefficient for both these variables was 0.74. Heart rate and rating of perceived exertion (RPE) were not significantly different between sessions (P = 0.42 and P = 0.095, respectively). The intra-class correlation coefficient for heart rate ranged from 0.65 to 0.97 and that for RPE from 0.85 to 0.91. We conclude that the 5-m multiple shuttle run test is a reliable measure of total and peak distances, heart rate and RPE response and is sufficiently reliabile to track changes in fitness over a season. The delta distance and fatigue index are not as reliable and should be interpreted with caution.     

Body sway,aim point fluctuation and performance in rifle shooters: inter- and intra-individual analysis

In this study, we examined the relationships between body sway, aim point fluctuation and performance in rifle shooting on an inter- and intra-individual basis. Six elite shooters performed 20 shots under competition conditions. For each shot, body sway parameters and four aim point fluctuation parameters were quantified for the time periods 5 s to shot, 3 s to shot and 1 s to shot. Three parameters were used to indicate performance. An AMTI LG6-4 force plate was used to measure body sway parameters, while a SCATT shooting analysis system was used to measure aim point fluctuation and shooting performance. Multiple regression analysis indicated that body sway was related to performance for four shooters. Also, body sway was related to aim point fluctuation for all shooters. These relationships were specific to the individual, with the strength of association, parameters of importance and time period of importance different for different shooters. Correlation analysis of significant regressions indicated that, as body sway increased, performance decreased and aim point fluctuation increased for most relationships. We conclude that body sway and aim point fluctuation are important in elite rifle shooting and performance errors are highly individual-specific at this standard. Individual analysis should be a priority when examining elite sports performance.     

The vibrational mode structure of a golf ball

In this paper, we report the discrete frequencies at which golf balls can vibrate, the mode patterns of these vibrations and how these modes can be excited. There are two broad classes of modes: those that radiate sound waves and those that do not. Both silent and acoustic modes are excited by tangential (i.e. spin-producing) impact forces; only acoustic modes are excited by radial impact forces. Exact analytical results for a homogeneous ball core are compared with finite element numerical results for both a core and a model two-piece ball. Correspondences are readily established for the important low-frequency modes, and the good agreement suggests the validity of these results for real golf balls. The results potentially provide the basis for a rapid, simple and non-destructive method of measuring the effective high-frequency elastic shear modulae of balls (and ball cores) as well as a method for 'tuning' the performance of balls for specific clubs. Some of these aspects are explored further in our companion paper in this issue.