An extreme learning machine (ELM) is a feedforward neural network (FNN) like learning system whose connections with output neurons are adjustable, while the connections with and within hidden neurons are randomly fixed. Numerous applications have demonstrated the feasibility and high efficiency of ELM-like systems. It has, however, been open if this is true for any general applications. In this two-part paper, we conduct a comprehensive feasibility analysis of ELM. In Part I, we provide an answer to the question by theoretically justifying the following: 1) for some suitable activation functions, such as polynomials, Nadaraya-Watson and sigmoid functions, the ELM-like systems can attain the theoretical generalization bound of the FNNs with all connections adjusted, i.e., they do not degrade the generalization capability of the FNNs even when the connections with and within hidden neurons are randomly fixed; 2) the number of hidden neurons needed for an ELM-like system to achieve the theoretical bound can be estimated; and 3) whenever the activation function is taken as polynomial, the deduced hidden layer output matrix is of full column-rank, therefore the generalized inverse technique can be efficiently applied to yield the solution of an ELM-like system, and, furthermore, for the nonpolynomial case, the Tikhonov regularization can be applied to guarantee the weak regularity while not sacrificing the generalization capability. In Part II, however, we reveal a different aspect of the feasibility of ELM: there also exists some activation functions, which makes the corresponding ELM degrade the generalization capability. The obtained results underlie the feasibility and efficiency of ELM-like systems, and yield various generalizations and improvements of the systems as well.

HIGHLIGHTSGenes linked to human psychiatric disorders can alter zebrafish collective behavior Differences from wildtype lead to ''scattered,'' ' 'coordinated,'' and ''huddled'' behavior Changes in individual interaction rules can explain emergent group level patterns

Multi-hop vehicle-to-vehicle communication is useful for supporting many vehicular applications that provide drivers with safety and convenience. Developing multi-hop communication in vehicular ad hoc networks (VANET) is a challenging problem due to the rapidly changing topology and frequent network disconnections, which cause failure or inefficiency in traditional ad hoc routing protocols. We propose an adaptive connectivity aware routing (ACAR) protocol that addresses these problems by adaptively selecting an optimal route with the best network transmission quality based on statistical and real-time density data that are gathered through an on-the-fly density collection process. The protocol consists of two parts: 1) select an optimal route, consisting of road segments, with the best estimated transmission quality, and 2) in each road segment of the chosen route, select the most efficient multi-hop path that will improve the delivery ratio and throughput. The optimal route is selected using our transmission quality model that takes into account vehicle densities and traffic light periods to estimate the probability of network connectivity and data delivery ratio for transmitting packets. Our simulation results show that the proposed ACAR protocol outperforms existing VANET routing protocols in terms of data delivery ratio, throughput and data packet delay. Since the proposed model is not constrained by network densities, the ACAR protocol is suitable for both daytime and nighttime city VANET scenarios.

No abstract

BackgroundA negative attitude toward disability is one of the potential barriers for people with disability (PWD) to achieve social equality. Although numerous studies have investigated attitudes toward disability, few have evaluated personal attitudes toward disability among PWD, and made comparisons with attitudes of healthy respondents. This study was to investigate and compare the attitudes of PWD, caregivers, and the public toward disability and PWD in China, to identify discrepancies in attitude among the three groupsand to examine potential influencing factors of attitude within each group.MethodsA cross-sectional study was conducted among 2912 PWD, 507 caregivers, and 354 members of the public in Guangzhou, China. Data were collected on participants’ socio-demographic information and personal attitudes toward disability using the Attitude to Disability Scale (ADS). ANOVA and ANCOVA were applied to compare the level of attitude among the three groups. Simple and multiple linear regression analyses were used to investigate the relationship between each background factor and attitude within each group.ResultsOver 90 % of caregivers were PWD’s family members. After controlling the socio-demographic characteristics, caregivers had the lowest total scores of ADS (caregivers: 47.7; PWD: 52.3; the public: 50.5). Caregivers who had taken care of PWD for longer durations of time had a more negative attitude toward disability. In contrast, PWD who had been disabled for longer times had a more positive attitude toward disability.ConclusionsThe current national social security system of China does not adequately support PWD’s family-member caregivers who may need assistance coping with their life with PWDs. More research is needed, and the development of a new health-care model for PWD is warranted.

The three versions of short-form WHOQOL-OLD contained the best items of the original module, much shorter, and with good internal consistency and criterion validity as a whole.

BackgroundUnder the circumstance of global population aging, the issue on how to facilitate the quality of life (QOL) for older people brings us grand challenge. On the way to solve this problem, it is inextricable to measure QOL for older people accurately at onset. This study is aimed at evaluating the reliability and validity of the Chinese version of the World Health Organization Quality of Life Instrument-Older Adults Module (WHOQOL-OLD).MethodsWe received 1005 valid WHOQOL-OLD questionnaires from 1050 respondents who were 60 and older by quota sampling method. To calculate the test-retest correlation coefficient we re-interviewed 101 participants from the community. Psychometric properties were evaluated from the aspect of feasibility, internal consistency reliability, test-retest reliability, content validity, construct validity and discriminant validity.ResultsMissing item responses took up 0.0%-2.7% in the scale. The WHOQOL-OLD showed satisfactory reliability with Cronbach’s Alpha coefficients ranging from 0.711 (Social participation) to 0.842 (Sensory ability) for each domain. The intra-class correlation coefficients (ICC) presenting test-retest reliability were all over 0.7. In Confirmatory Factor Analysis (CFA), Root Mean Square Error of Approximation (RMSEA) was 0.084 (a little more than 0.08) and comparative fit index (CFI) 0.95 (>0.90) which meant acceptable construct validity. There were higher correlation coefficients between items and their hypothesized domains than other domains (P < 0.001), indicating good content validity. The results of t-test showed good discriminant validity of the WHOQOL-OLD between the healthy group and the unhealthy group (P < 0.0083).ConclusionThe Chinese version of WHOQOL-OLD showed good feasibility, reliability and validity in this study. However, before it can be used national-widely, further research should be conducted in other areas of China.

Abstract-An extreme learning machine (ELM) can be regarded as a two stage feed-forward neural network (FNN) learning system which randomly assigns the connections with and within hidden neurons in the first stage and tunes the connections with output neurons in the second stage. Therefore, ELM training is essentially a linear learning problem, which significantly reduces the computational burden. Numerous applications show that such a computation burden reduction does not degrade the generalization capability. It has, however, been open that whether this is true in theory. The aim of our work is to study the theoretical feasibility of ELM by analyzing the pros and cons of ELM. In the previous part on this topic, we pointed out that via appropriate selection of the activation function, ELM does not degrade the generalization capability in the expectation sense. In this paper, we launch the study in a different direction and show that the randomness of ELM also leads to certain negative consequences. On one hand, we find that the randomness causes an additional uncertainty problem of ELM, both in approximation and learning. On the other hand, we theoretically justify that there also exists an activation function such that the corresponding ELM degrades the generalization capability. In particular, we prove that the generalization capability of ELM with Gaussian kernel is essentially worse than that of FNN with Gaussian kernel. To facilitate the use of ELM, we also provide a remedy to such a degradation. We find that the well-developed coefficient regularization technique can essentially improve the generalization capability. The obtained results reveal the essential characteristic of ELM and give theoretical guidance concerning how to use ELM.