Optimal metric selection for improved multi-pose face recognition with group information

We address the limitation of sparse representation based classification with group information for multi-pose face recognition. First, we observe that the key issue of such classification problem lies in the choice of the metric norm of the residual vectors, which represent the fitness of each class. Then we point out that limitation of the current sparse representation classification algorithms is the wrong choice of the ℓ2 norm, which does not match with data statistics as these residual values may be considerably non-Gaussian. We propose an explicit but effective solution using ℓp norm and explain theoretically and numerically why such metric norm would be able to suppress outliers and thus can significantly improve classification performance comparable to the state-of-arts algorithms on some challenging datasets

Learning sparse latent representation and distance metric for image retrieval

The performance of image retrieval depends critically on the semantic representation and the distance function used to estimate the similarity of two images. A good representation should integrate multiple visual and textual (e.g., tag) features and offer a step closer to the true semantics of interest (e.g., concepts). As the distance function operates on the representation, they are interdependent, and thus should be addressed at the same time. We propose a probabilistic solution to learn both the representation from multiple feature types and modalities and the distance metric from data. The learning is regularised so that the learned representation and information-theoretic metric will (i) preserve the regularities of the visual/textual spaces, (ii) enhance structured sparsity, (iii) encourage small intra-concept distances, and (iv) keep inter-concept images separated. We demonstrate the capacity of our method on the NUS-WIDE data. For the well-studied 13 animal subset, our method outperforms state-of-the-art rivals. On the subset of single-concept images, we gain 79:5% improvement over the standard nearest neighbours approach on the MAP score, and 45.7% on the NDCG.

Face recognition from face motion manifolds using robust kernel resistor-average distance

In this work we consider face recognition from face motion manifolds. An information-theoretic approach with Resistor-Average Distance (RAD) as a dissimilarity measure between distributions of face images is proposed. We introduce a kernel-based algorithm that retains the simplicity of the closed-form expression for the RAD between two normal distributions, while allowing for modelling of complex, nonlinear manifolds. Additionally, it is shown how errors in the face registration process can be modelled to significantly improve recognition. Recognition performance of our method is experimentally demonstrated and shown to outperform state-of-the-art algorithms. Recognition rates of 97–100% are consistently achieved on databases of 35– 90 people.

Comfort response of three silicone hydrogel daily disposable contact lenses

To evaluate subjective ocular comfort across the day with three silicone hydrogel daily disposables (SHDDs) in a group of adapted lens wearers. Masked subjects (asymptomatic or symptomatic of end-of-day (EOD) dryness with habitual lenses) wore three SHDDs: DAILIES TOTAL1 (DT1), Clariti 1day (C1D), or 1-DAY ACUVUE TRUEYE (AVTE), each for 3 days. On day 2, wearing time (WT) and comfort ratings after insertion, at 4, 8, and 12 hours, and at EOD were recorded. Because not all subjects wore lenses for 12 hours, comfort was analyzed across the day (up to 8 hours, 8 to 12 hours), and a new variable (“cumulative comfort” [CC]) was calculated for EOD. One hundred four subjects completed the study (51 asymptomatic, 53 symptomatic). The two groups had different WTs (mean WT, 14.0 and 12.7 hours, respectively; p < 0.001). Ocular comfort was rated higher in the asymptomatic group throughout the day (p < 0.001). One hundred four subjects wore all three SHDDs for at least 8 hours, whereas 74 (45 asymptomatic, 29 symptomatic) subjects wore them for 12 hours or longer. Comfort ratings were higher with DT1 (least square means [LSM] = 91.0) than with C1D (LSM = 86.5; p < 0.001) and AVTE (LSM = 87.7; p = 0.011) for the first 8 hours and lower with C1D compared with DT1 (p = 0.012) from 8 to 12 hours. Mean EOD (± SD) comfort with the C1D lens was 72 ± 21, lower than both DT1 (mean, 79 ± 17; p = 0.001) and AVTE (mean, 78 ± 21; p = 0.010). Mean CC was higher in the asymptomatic group (mean, 1261 ± 59) compared with that in the symptomatic group (mean, 1009 ± 58; p < 0.001) and higher for DT1 (mean, 1184 ± 258) than C1D (mean, 1094 ± 318; p = 0.002) and AVTE (mean, 1122 ± 297; p = 0.046). All three SHDDs had average WTs of 12 hours or longer for 1 day. Comfort during the first 12 hours was highest with DT1 (similar to AVTE between 8 and 12 hours) and lowest with C1D. End-of-day comfort was lowest with C1D, and CC was highest for DT1. Cumulative comfort may be a valuable new metric to assess ocular comfort during the day.

The influence of rare earth mercaptoacetate on the initiation of corrosion on AA2024-T3 Part I: average statistics of each intermetallic composition

De-alloying of S-phase in AA2024-T3 in the presence chlorides, is well-known. However, it is unclear how rare earth mercaptoacetate inhibitors affect this process when immersed in a 0.1. M NaCl solution. This paper analyses data obtained using EPMA on AA2024-T3 surfaces before and after a 16. min immersion period. Cerium and praseodymium mercaptoacetate inhibited the de-alloying process of S-phase particles. Although no significant change in composition was observed for cathodic intermetallics, each appeared to participate in local corrosion reactions as evidenced by the development of surface oxides. Clustering between S-phase and one of the Cu-containing intermetallic domains was also evident.

Habitat switching by Bewick's swans: Maximization of average long-term energy gain?

1. In a system where depletion drives a habitat shift, the hypothesis was tested that animals switch habitat as soon as the average daily net energy intake (or gain) drops below that attainable in the alternative habitat.

2. The study was performed in the Lauwersmeer area. Upon arrival during the autumn migration, Bewick's swans first feed on below-ground tubers of fennel pondweed on the lake, but subsequently switched to feeding on harvest remains in sugar beet fields.

3. The daily energy intake was estimated by multiplying the average time spent foraging per day with the instantaneous energy intake rate while foraging. In the case of pondweed feeding, the latter was estimated from the functional response and the depletion of tuber biomass. In the case of beet feeding, it was estimated from dropping production rate. Gross energy intake was converted to metabolizable energy intake using the assimilation as determined in digestion trials. The daily energy expenditure was estimated by the time-energy budget method. Energetic costs were determined using heart rate.

4. The daily gain of pondweed feeding at the median date of the habitat switch (i.e. when 50% of the swans had switched) was compared with that of beet feeding. The daily gain of beet feeding was calculated for two strategies depending on the night activity on the lake: additional pondweed feeding (mixed feeding) or sleeping (pure beet feeding).

5. The majority of the swans switched when the daily gain they could achieve by staying on the pondweed bed fell just below the average daily gain of pure beet feeders. However, mixed feeders would attain an average daily gain considerably above that of pondweed feeders. A sensitivity analysis showed that this result was robust.

6. We therefore reject the hypothesis that the habitat switch by swans can be explained by simple long-term energy rate maximization. State-dependency, predation risk, and protein requirements are put forward as explanations for the delay in habitat switch.

Effect of mean stress on multiaxial ratcheting life: a simplified life prediction model based on average equivalent stress amplitude

This article proposes a model to predict uniaxial and multiaxial ratcheting life by addressing the three primary parameters that influence failure life: fatigue damage, ratcheting damage and the multiaxial loading path. These three factors are addressed in the present model by (a) the stress amplitude for fatigue damage, (b) mean stress-dependent Goodman equation for ratcheting damage and (c) an inherent weight factor based on average equivalent stress to account for the multiaxial loading. The proposed model requires only two material constants which can be easily determined from uniaxial symmetric stress-controlled fatigue tests. Experimental ratcheting life data collected from the literature for 1025 and 42CrMo steel under multiaxial proportional and nonproportional constant amplitude loading ratcheting with triangular sinusoidal and trapezoidal waveform (i.e. linear, rhombic, circular, elliptical and square stress paths) have shown good agreement with the proposed model.