KWTA networks and their applications

Winner-Take-All (WTA) or K-Winner-Take-All (KWTA) networks have been frequently used as the basic building blocks of complex neural networks. This paper introduces a new selection rule for network connections that implements stable KWTA networks. To widen the applications of WTA networks, a new class of WTA networks is proposed, and their efficient design methods are presented. We demonstrate the properties of the generalized class of WTA networks, through three application examples.     

Bone growth and bone development in the presence of implants or after induced leg-lengthening studied using the Oxford Scanning Proton Microprobe

To respond to varying environmental demands the bone tissue in the body is under continual reconstruction throughout life. It is known that metallic elements are important for maintaining normal bone structure, but their roles are not well understood. More information about the effects of metal excess or deficiency is needed to help in the development of metallic bone implants and to improve the treatment of bone fractures and defects. The Oxford Scanning Proton Microprobe (SPM) is being applied in two studies involving metal ions in bone: (1) bone regrowth and bonding to titanium bone implants may be influenced by diffusion of Ti ions into the bone. We are using microPIXE to determine the metal ion content of bone developing in contact with implants of pure Nb, Ti and Ti alloys. (2) Bone lengthening as a surgical procedure is induced by fracturing the bone and allowing it to heal with a small gap between the fractured ends created by the use of external fixators. The gap can be slowly increased during the healing process to stimulate the production of new bone. The enzymes and other constituents of the developing bone need certain metals for their function. Using experimental animals we have studied the concentrations of the metals and whether a deficiency of trace metals limits the optimum rate of bone lengthening.     

TiO_2对ZrO_2－NiCr金属陶瓷烧结特性和相稳定性的影响

本文研究了ＴｉＯ２掺入对ＣａＯ稳定ＺｒＯ２（简称ＣａＳＺ－ＮｉＣｒ）金属陶瓷烧结特性的影响，测定了ＣａＳＺ－ＮｉＣｒ－ＴｉＯ２金属陶瓷烧结样品的密度和显微硬度，结果表明，ＴｉＯ２的掺入能提高ＣａＳＺ－ＮＩＣｒ金属陶瓷的烧结密度和显微硬度，对ＣａＳＺ－ＮｉＣｒ金属陶瓷有助烧结作用．金相观察表明，ＴｉＯ３的掺入改变了ＮｉＣｒ相在金属陶瓷中的形貌，说明了ＮｉＣｒ与ＣａＳＺ的浸润性有一定的改善．ＸＲＤ和ＥＰＭＡ分析结果显示，ＴｉＯ２偏聚在ＮｉＣｒ金属相中，不影响ＣａＳＺ的稳定性．     

Limited evidence of zebra mussel (Dreissena polymorpha) consumption by freshwater drum (Aplodinotus grunniens) in Lake Winnipeg

Freshwater drum (Aplodinotus grunniens) may be a predator of the invasive zebra mussel (Dreissena polymorpha), which established in Lake Winnipeg in 2013. In this study, the diets, trophic position, and growth of 51 freshwater drum collected in 2019 (six years post-zebra mussel invasion) were compared to 64 freshwater drum sampled in 2000. Benthic insect larvae were the dominant food items in both years. Although mollusks occur in high densities in Lake Winnipeg, they were only consumed by a few freshwater drum in either sample year. Zebra mussels were not a frequent prey item in 2019 as they were only consumed by four of the sampled freshwater drum. Stable isotope analysis of white muscle tissue yielded similar δ¹³C and δ¹⁵N values in both years and were consistent with a benthic, insectivorous diet. Length-at-age data derived from otoliths revealed that the 2019 population had at least an equal growth rate to the 2000 population. Weight-at-length data suggested that fish condition was greater in 2019 than in 2000, which coincided with increased benthic macroinvertebrate density in Lake Winnipeg. Based on these findings, Lake Winnipeg freshwater drum continue to feed predominantly on insect larvae and not zebra mussels.     

Spatial ecology of common carp (Cyprinus carpio) in Lake Winnipeg and its potential for management actions

Common carp feeding and spawning behaviours negatively impact the functioning of marsh ecosystems. In the Netley-Libau Marsh, situated on the southern end of Lake Winnipeg, water level regulations, nonpoint source nutrient pollution, and the non-native common carp are thought to be the main contributors to the degradation of the marsh habitat. Using acoustic telemetry, we analysed the movement rate, frequency and timing of suspected spawning migrations, spatial ecology, and aggregation of common carp in the Lake Winnipeg drainage over a three year time period. Common carp moved the farthest during the open water period when water temperature was >5 °C. Their annual migration into Netley-Libau Marsh was correlated to ordinal date. Common carp left the marsh in late spring/early summer, presumably feeding in Lake Winnipeg, before moving to overwintering sites situated in Traverse Bay and Lake Winnipeg, where they arrived typically by October and formed aggregations. These findings will inform habitat and fisheries managers in the effort to undertake evidence-based management actions. The predictability of the movements and the tendency for common carp to aggregate indicates that exclusion techniques and commercial fishing may represent viable management solutions.     

Developing habitat associations for fishes in Lake Winnipeg by linking large scale bathymetric and substrate data with fish telemetry detections

Understanding relationships between freshwater fishes and habitat is critical for effective fisheries and habitat management. Habitat suitability indices (HSI) are commonly used to describe fish–habitat associations in rivers and other freshwater ecosystems. When applied to large lakes however, standard sampling procedures are inadequate because of larger sampling areas and an increased risk of fish collection bias through one-time observations. Here, we use lake bathymetry, substrate, and multiple fish telemetry detections collected from a systematically deployed receiver grid to develop HSI for four fish species (lake sturgeon, freshwater drum, common carp, and walleye) in Lake Winnipeg. Seasonal variations in habitat use based on water depth and substrate were observed in three of four species. Lake sturgeon remained in shallow locations with predominantly gravel substrate near the mouth of the Winnipeg River regardless of season. Freshwater drum persisted over fine substrate in both summer and winter but had a broader depth range in the summer compared to winter. Common carp shifted from mid-range depths and silt substrate in the summer to shallow depths and gravel substrate in the winter. Walleye showed an unchanging association to fine substrate but expanded from primarily mid-range depths in the summer to include shallower depths in the winter. These findings show how multiple telemetry detections per fish can be combined with hydroacoustic data to provide informative habitat associations for fishes in a large lacustrine ecosystem.     

Semisupervised Encrypted Traffic Identification Based on Auxiliary Classification Generative Adversarial Network

The rapidly increasing popularity of mobile devices has changed the methods with which people access various network services and increased network traffic markedly. Over the past few decades, network traffic identification has been a research hotspot in the field of network management and security monitoring. However, as more network services use encryption technology, network traffic identification faces many challenges. Although classic machine learning methods can solve many problems that cannot be solved by port- and payload-based methods, manually extract features that are frequently updated is time-consuming and labor-intensive. Deep learning has good automatic feature learning capabilities and is an ideal method for network traffic identification, particularly encrypted traffic identification; Existing recognition methods based on deep learning primarily use supervised learning methods and rely on many labeled samples. However, in real scenarios, labeled samples are often difficult to obtain. This paper adjusts the structure of the auxiliary classification generation adversarial network (ACGAN) so that it can use unlabeled samples for training, and use the wasserstein distance instead of the original cross entropy as the loss function to achieve semisupervised learning. Experimental results show that the identification accuracy of ISCX and USTC data sets using the proposed method yields markedly better performance when the number of labeled samples is small compared to that of convolutional neural network (CNN) based classifier.     

Parallel source separation system for heart and lung sounds

The Journal of Supercomputing - In this paper, we propose a parallel source separation system designed to extract heart and lung sounds from single-channel mixtures. The proposed system is based on...