A hybrid knowledge and ensemble classification approach for prediction of venous thromboembolism

Clinical narratives such as progress summaries, lab reports, surgical reports, and other narrative texts contain key biomarkers about a patient's health. Evidence-based preventive medicine needs accurate semantic and sentiment analysis to extract and classify medical features as the input to appropriate machine learning classifiers. However, the traditional approach of using single classifiers is limited by the need for dimensionality reduction techniques, statistical feature correlation, a faster learning rate, and the lack of consideration of the semantic relations among features. Hence, extracting semantic and sentiment-based features from clinical text and combining multiple classifiers to create an ensemble intelligent system overcomes many limitations and provides a more robust prediction outcome. The selection of an appropriate approach and its interparameter dependency becomes key for the success of the ensemble method. This paper proposes a hybrid knowledge and ensemble learning framework for prediction of venous thromboembolism (VTE) diagnosis consisting of the following components: a VTE ontology, semantic extraction and sentiment assessment of risk factor framework, and an ensemble classifier. Therefore, a component-based analysis approach was adopted for evaluation using a data set of 250 clinical narratives where knowledge and ensemble achieved the following results with and without semantic extraction and sentiment assessment of risk factor, respectively: a precision of 81.8% and 62.9%, a recall of 81.8% and 57.6%, an F measure of 81.8% and 53.8%, and a receiving operating characteristic of 80.1% and 58.5% in identifying cases of VTE.     

Distributed Learning for Planning Under Uncertainty Problems with Heterogeneous Teams

This paper considers the problem of multiagent sequential decision making under uncertainty and incomplete knowledge of the state transition model. A distributed learning framework, where each agent learns an individual model and shares the results with the team, is proposed. The challenges associated with this approach include choosing the model representation for each agent and how to effectively share these representations under limited communication. A decentralized extension of the model learning scheme based on the Incremental Feature Dependency Discovery (Dec-iFDD) is presented to address the distributed learning problem. The representation selection problem is solved by leveraging iFDD’s property of adjusting the model complexity based on the observed data. The model sharing problem is addressed by having each agent rank the features of their representation based on the model reduction error and broadcast the most relevant features to their teammates. The algorithm is tested on the multi-agent block building and the persistent search and track missions. The results show that the proposed distributed learning scheme is particularly useful in heterogeneous learning setting, where each agent learns significantly different models. We show through large-scale planning under uncertainty simulations and flight experiments with state-dependent actuator and fuel-burn- rate uncertainty that our planning approach can outperform planners that do not account for heterogeneity between agents.     

Use of a simulation game in delivering blended lifelong learning in the construction industry – Opportunities and Challenges

Continuing professional development (CPD) and life-long learning are vital to both individual and organisational success. For higher education, the intensive resource requirements requisite in the development of e-learning content and the challenges in accommodating different learning styles, developing an e-learning program can be a resource intensive exercise. A blended learning program has been developed in Ireland in an attempt to address the CPD needs of Irish construction professionals. This initiative attempts to strike a balance between the considerable resources required in the development of an e-learning initiative while addressing staff concerns in integrating technology in the delivery of programs. This pilot explores the issues encountered in integrating a simulation game, called MERIT, on a module as part of this blended learning program. The key finding from this research indicates that simulation games can play a very effective role in the delivery of lifelong learning opportunities aimed at the construction industry. However, management of the integration of a simulation game into a program requires careful planning, establishing key milestone dates and encouraging online collaboration through assigning of marks for effort and use of voice over IP communications.     

Query-based learning for aerospace applications

Models of real-world applications often include a large number of parameters with a wide dynamic range, which contributes to the difficulties of neural network training. Creating the training data set for such applications becomes costly, if not impossible. In order to overcome the challenge, one can employ an active learning technique known as query-based learning (QBL) to add performance-critical data to the training set during the learning phase, thereby efficiently improving the overall learning/generalization. The performance-critical data can be obtained using an inverse mapping called network inversion (discrete network inversion and continuous network inversion) followed by oracle query. This paper investigates the use of both inversion techniques for QBL learning, and introduces an original heuristic to select the inversion target values for continuous network inversion method. Efficiency and generalization was further enhanced by employing node decoupled extended Kalman filter (NDEKF) training and a causality index (CI) as a means to reduce the input search dimensionality. The benefits of the overall QBL approach are experimentally demonstrated in two aerospace applications: a classification problem with large input space and a control distribution problem.     

L'acoustique des salles

Due to the lack of effective design tools, the Acoustic Department of CSTB in France has developed a computed method to allow the designer to control the main acoustic parameters during design planning and improve the reliability of the results. Called the ‘conical beam method’, it is based on geometrical acoustics and makes it possible to calculate, for any room and any point, the room impulse response, from which the main objective acoustic criteria are derived. A simulation of the actual sound in the room can then be made before construction.     

Age differences in conversational source monitoring.

The present investigation simulated a group conversation in which participants asked (inquirer) and answered (responder) questions, as well as listened to others exchange information. Source (inquirer; responder) identification accuracy was evaluated immediately or after 1 week. Older adults were less adept at source identification, although this difference was reduced with personal (Experiment 2) rather than categorical (Experiment 1) topics. The age difference was independent of explicit memory (cued recall and recognition), suggesting that memory for source and information are separable. Older adults were comparable to younger adults in responder identification but worse at inquirer identification. Responder identification was better than inquirer identification, with the latter dropping to chance at 1 week. Source identification was most accurate when participants were in the responder role; there was little difference between the inquirer and listener roles. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

The Interface of TiB<Subscript>2</Subscript> and Al<Subscript>3</Subscript>Ti in Molten Aluminum

In the grain refinement of aluminum, Al₃Ti and TiB₂ particles are introduced to reduce the casting grain size down to 200 micrometer level, which makes cold working possible. The particles are brought in by the addition of Al-Ti-B-type master alloys. It is generally believed that TiB₂ particles are stable and nucleate α-Al grains in solidification in the presence of titanium in solution from the dissolution of Al₃Ti particles in the master alloys. The titanium in solution either forms Al₃Ti layers on the surface of TiB₂ particles to promote the nucleation of α-Al grains or remains as solute to restrict the growth of α-Al grains in solidification. However, a consensus on a grain refinement mechanism is still to be reached due to the lack of direct observation of the three phases in castings. This paper presents finding of the TiB₂/Al₃Ti interfaces in an Al-Ti-B master alloy. It demonstrates a strong epitaxial growth of Al₃Ti on the surface of TiB₂ particles, a sign of the formation of an Al₃Ti layer on the surface of TiB₂ particles in grain refinement practice. The Al₃Ti layer has a crystal coherency with α-Al and hence offers a substrate for heterogeneous nucleation of α-Al grains. However, the layer must be dynamic to avoid the formation of compounded Al₃Ti and TiB₂ particles leading to the loss of efficiency in grain refinement.     

A highly conserved intraspecies homolog of the Saccharomyces cerevisiae elongation factor-3 encoded by the HEF3 gene

A paralog (intraspecies homolog) of the Saccharomyces cerevisiae YEF3 gene, encoding elongation factor-3, has been sequenced in the course of the yeast genome project, and identified by database searching; this gene has been designated HEF3. Bioinformatic and Northern blot analysis indicate that the HEF3 gene is not expressed during vegetative growth. Deletion of the HEF3 gene reveals no growth defects, nor any defects in mating or sporulation. A high copy 2μ clone of HEF3 was constructed, and was shown to be unable to complement a null allele of yef3. Finally, an in vitro assay for ribosome-stimulated ATPase activity was performed with isogenic HEF3 and Δhef3 strains; no difference in biochemical activity could be detected in these strains. From these results, we conclude that the HEF3 gene does not encode a functional homolog of YEF3. © 1998 John Wiley & Sons, Ltd.     

Combined extraction processes of lipid from Chlorella vulgaris microalgae: microwave prior to supercritical carbon dioxide extraction

Extraction yields and fatty acid profiles from freeze-dried Chlorella vulgaris by microwave pretreatment followed by supercritical carbon dioxide (MW-SCCO(2)) extraction were compared with those obtained by supercritical carbon dioxide extraction alone (SCCO(2)). Work performed with pressure range of 20-28 Mpa and temperature interval of 40-70 °C, gave the highest extraction yield (w/w dry weight) at 28 MPa/40 °C. MW-SCCO(2) allowed to obtain the highest extraction yield (4.73%) compared to SCCO(2) extraction alone (1.81%). Qualitative and quantitative analyses of microalgae oil showed that palmitic, oleic, linoleic and α-linolenic acid were the most abundant identified fatty acids. Oils obtained by MW-SCCO(2) extraction had the highest concentrations of fatty acids compared to SCCO(2) extraction without pretreatment. Native form, and microwave pretreated and untreated microalgae were observed by scanning electronic microscopy (SEM). SEM micrographs of pretreated microalgae present tearing wall agglomerates. After SCCO(2), microwave pretreated microalgae presented several micro cracks; while native form microalgae wall was slightly damaged.