Extracting Hidden Context

Concept drift due to hidden changes in context complicates learning in many domains including financial prediction, medical diagnosis, and communication network performance. Existing machine learning approaches to this problem use an incremental learning, on-line paradigm. Batch, off-line learners tend to be ineffective in domains with hidden changes in context as they assume that the training set is homogeneous. An off-line, meta-learning approach for the identification of hidden context is presented. The new approach uses an existing batch learner and the process of contextual clustering to identify stable hidden contexts and the associated context specific, locally stable concepts. The approach is broadly applicable to the extraction of context reflected in time and spatial attributes. Several algorithms for the approach are presented and evaluated. A successful application of the approach to a complex flight simulator control task is also presented.     

电场对Al-Mn-Mg合金的微结构和织构形成的影响

利用透射电子显微镜观察(TEM)和X射线衍射技术(ODFs分析),研究了电场对Al-Mn-Mg合金的回复和再结晶组织演变、再结晶织构的形成和发展的影响.结果表明,电场对再结晶的影响其强度有一个门槛值(3～4 kV/cm),低于此值电场对该合金的再结晶没有明显的影响.强度为4 kV/cm的电场对再结晶形核的影响较大,可抑制Al-Mn-Mg合金的回复和再结晶,促进再结晶立方织构的形成.其主要原因是电场降低了各取向的形变储能,推迟了再结晶进程,抑制储存能小的取向晶核的形成和长大,促进储存能大的S取向晶粒向立方织构择优生长.     

Book reviews

Abstract

When news of the assassination of Mohandas Gandhi circulated through New Delhi on 30 January 1948, Margaret Bourke-White, one of Life magazine's premier photographers, and Henri Cartier-Bresson, then a comparatively unknown French photojournalist, raced to Birla House, where the event occurred. Both understood the journalistic imperative of photographing Gandhi's body. Because of their different philosophical and technical approaches to photography, they responded to the challenge in different ways and produced very different results. This study compares their ideas, approaches, and results. It also offers a case study of one of Cartier-Bresson' s most important news reportages.¹     

Reducing the power of wireless terminals by adaptive baseband processing

In this paper, we illustrate specific power savings obtained from exploiting a reconfigurable mobile terminal under the 3GPP LTE standard. Building on traditional link adaptation towards maximum throughput and extended towards minimal power consumption, we add two flexible baseband components: the turbo decoder and the multiple input multiple output (MIMO) detector. Optimizing their configuration leads to larger power savings when compared to non-flexible systems only performing link adaptation. The gain observed strongly depends on the scenario. For low-activity set-ups with a few minutes of voice per day, the idle power dominates and the active data rate is relatively low. This makes analog front-end and time-domain processing dominant given their constant power consumption while MIMO detection and turbo decoding that scale with data rate play a smaller role. Still, because of its ability to improve the system spectral efficiency and hence reduce its duty cycle, an advanced MIMO detector can save 10% in power consumption, on the condition that the network requires to use MIMO. Otherwise single input single output is more power-efficient in downlink. In high-throughput scenarios, larger gains are obtained. The flexible MIMO detector can save up to 35% of average power consumption. The turbo decoding also brings some gain, saving up to 12% of power when the full bandwidth is allocated to a single user.     

Optical and electrical characterization of carbon nanoparticles produced in laminar premixed flames

Optical and electrical properties of carbonaceous particles produced in laboratory scale, premixed ethylene/air flames are obtained. Light absorption and Raman spectroscopy show that the change in particle nanostructure follows a graphitization trajectory as the flame richness increases. The optical band gap decreases and the size of the aromatic network in the particle increases, while the interlayer spacing between parallel layers decreases. The electrical conductivity of the materials increases by increasing flame richness in agreement to the graphitization trajectory. A non-ohmic behavior has been found and explained in terms of electron tunneling in a percolative network. Our results show that the electrical properties of flame formed carbon nanoparticles are strongly dependent on their nanostructure, and hence they have to be used carefully for the determination of particle concentration with conductometric sensors. Moreover, the dependence of the electrical properties of combustion formed particles might be useful for the development of cheap sensors for the selective detection of different classes of combustion aerosols.