5916 research outputs found
Sort by
Combined weight and density bounds on the polynomial threshold function representation of Boolean functions
No full textIn an earlier report it was shown that an arbitrary n-variable Boolean function f can be represented as a polynomial threshold function (PTF) with 0.75×2n or less number of monomials. In this report, we derive an upper bound on the absolute value of the (integer) weights of a PTF that represents f and still obeys the aforementioned density bound. To our knowledge this provides the best combined bound on the PTF density (number of monomials) and PTF weight (sum of the coefficient magnitudes) of general Boolean functions. For the special case of bent functions, it is found that any n-variable bent function can be represented with integer coefficients less than or equal to 2n with density no more than 0.75×2n, and for the case of m-sparse Boolean functions that are almost constant except for small (m≪2n) number of variable assignments, it is shown that they can be represented with small weight PTFs with density at most m+2n−1. In addition, tight PTF weight bounds with conformance to the density bound of 0.75×2n are numerically obtained for the general Boolean functions up to 6 variables.Osaka Universit
Reflector-aided underwater optical channel modeling
No full textUnderwater visible light communication (UVLC) has been proposed as a high-speed alternative to acoustic signaling. While most UVLC systems are configured to work in line-of-sight (LOS) conditions, it is also possible to exploit reflected signals for performance enhancements. In this Letter, we propose a closed-form expression for the underwater path loss assuming non-LOS (NLOS) transmission through the water surface and man-made reflector (e.g., mirror) in addition to the LOS link. Utilizing the derived expression, we quantify the achievable NLOS gain defined as the ratio between the maximum achievable channel coefficient from reflection and the overall channel coefficient. We validate our findings experimentally by utilizing the water surface and the mirror as the reflecting surfaces in an aquarium. Our results reveal that achievable gains up to around 3 dB can be observed due to reflections
On existence of equilibrium under social coalition structures
No full textIn a strategic-form game, a strategy profile is an equilibrium if no viable coalition of agents (or players) benefits (in the Pareto sense) from jointly changing their strategies. Weaker or stronger equilibrium notions can be defined by considering various restrictions on coalition formation. For instance, in a Nash equilibrium, it is assumed that viable coalitions are singletons, and in a super strong equilibrium, it is assumed that every coalition is viable. Restrictions on coalition formation can be justified by communication limitations, coordination problems, or institutional constraints. In this paper, inspired by social structures in various real-life scenarios, we introduce certain restrictions on coalition formation, and on their basis, we introduce a number of equilibrium notions. As an application, we study our equilibrium notions in resource selection games (RSGs), and we present a complete set of existence and nonexistence results for general RSGs and their important special cases.TÜBİTA
Review of uncertainties in building characterization for urban-scale energy modeling
No full textBottom-up modeling appears to be a suitable approach for the urban-scale building energy performance assessment with providing valuable inferences on the complicated building energy patterns and helping authorities monitor/predict the energy demand for urban planning and retrofitting. Archetype characterization is the utmost challenging process when developing bottom-up models since there is a large diversity in characteristic features of building stocks. This gap induces practitioners to seek stochastic methods even though the deterministic approaches are solid guides in archetype characterization. Hence, the research objective of this study is to provide insights into the motivation, challenges, and methods of the studies conducted to assess the buildings' energy demand at the urban scale. The original value of this research is to analyze/question different archetype characterization methods and their practicability over wide-ranging studies, identify the most crucial characterization parameters and assess the validation techniques to enhance the demand estimations of urban building energy models (UBEMs). To that end, this study performs a literature review and mainly provides the following findings: (1) The required characterization method is highly dependent on the purpose and scope of the study. (2) The Bayesian calibration makes ground in UBEM practices as it consolidates the models' estimation power through the probabilistic archetype characterization. (3) Considering the notable fluctuations in buildings' energy demand induced by occupancy patterns, detailed occupancy profiles could improve the archetype characterization. Finally, the major setback is the lack of available data to characterize energy models with building-specific information. (4) Building information models (BIMs) could soon play a pivotal role in supplying such data for UBEM practices. This study contributes to the literature by fulfilling the lack of perspective that concentrates on the archetype characterization methods in UBEM. The findings could help practitioners (e.g., policymakers and city planners) and academics to comprehend the potential of the UBEM that improves energy management strategies at the urban scale
Experimental impact analysis of the refrigerator cable distributions on conducted electromagnetic emission
No full textAs the number of electronic devices in our daily lives increases day by day, the risk of electromagnetic interference (EMI) between the devices and between the device and the grid increases. Refrigerators are the only devices found in almost every home and work non-stop. For this reason, refrigerators have importance in terms of electromagnetic interference. Three refrigerators with different cable distributions have been prepared for this research. Conducted emission tests have been performed for phase lines and neutral lines to examine the impact of the cable distributions on EMI. In some cases, separation of AC and DC cables improved the noise values by around 2-4 dB at most of the frequencies; however, in some cases, the results were so worse that the product failed the test
To compete or not compete: Contributions of children’s regulation and gender to their competitive behaviors
No full textPreschool children naturally display competitive behavioral patterns. The purpose of the current study was to investigate the association between preschool children’s regulation (regulatory and control components) and competitive behaviors (task-oriented and other-referenced). A total of 260 preschool children (47.7% girls) ranging in age from 49 to 72 months (M = 63.83, SD = 6.17) were recruited for the current study. The participating teachers reported on children’s regulation and competitive behaviors. Hierarchical multiple regression analyses that accounted for the nesting structure of the data revealed that children’s regulation and control skills were significantly related to their task-oriented competition. Child gender moderated the association between regulation and task-oriented competition such that being highly regulated contributed to children’s task-oriented competition, specifically for boys. Control skills were negatively associated with children’s other-referenced competition. Implications of the study and future directions are discussed
Relief aid provision to en route refugees: Multi-period mobile facility location with mobile demand
No full textMany humanitarian organizations aid en route refugee groups who are on their journey to cross borders using mobile facilities and need to decide the number and routes of the facilities. We define a multi-period facility location problem in which both the facilities and demand are mobile on a network. Refugee groups may enter and exit the network in different periods and follow various paths. In each period, a refugee group moves from one node to an adjacent one in their predetermined path. Each facility should be located at a node in each period and provides service to the refugees at that node. Each refugee should be served at least once in a predetermined number of consecutive periods. The problem is to locate the facilities in each period to minimize the total setup and travel costs of the mobile facilities, while ensuring the service requirement. We call this problem the multi-period mobile facility location problem with mobile demand (MM-FLP-MD) and prove its NP-hardness. We formulate a mixed integer linear programming (MILP) model and develop an adaptive large neighborhood search algorithm (ALNS) to solve large-size instances. We tested the computational performance of the MILP and the metaheuristic algorithm by extracting data from the 2018 Honduras Migration Crisis. For instances solved to optimality by the MILP model, the proposed ALNS determines the optimal solutions faster and provides better solutions for the remaining instances. By analyzing the sensitivity to different parameters, we provide insights to decision-makers.TÜBİTA
Kısıtlandırılmamış motor sensör öğrenimi ile üst seviye temsiller
No full textLiving organisms, in particular mammals are adept at learning complex tasks that may require basic planning such as tool use and manipulation. This ability is manifested by the central nervous system; in particular by the brain, which has evolved from neural circuits that perform low level signal processing for action and perception. Artificial systems can learn to perform low level signal processing for tasks requiring basic perception and action. A particular learning mechanism for this is Reinforcement Learning (RL), in which the sensorimotor experience of an agent that involves a sequence of action decisions and returned 'rewards' from the environment are used to build a 'policy' for a given task. The complex task execution capabilities of higher-level animals may be facilitated by the abstractions formed based on the sensorimotor life cycle of lower-level neural circuits during learning. In a similar vein, the learning experience of an RL agent can be used to form symbol-like structures to enable more complex task executions, by for example explicit planning in symbolic space. The emergence of such high-level structures and their relationship with the environment and the agent dynamics may be revealed by analyzing the neural outputs during learning and actual deployment in different environments. In the literature, there are a range of studies that start by assuming the existence of symbolic structures that 'ground' them onto continuous sensorimotor signals experienced by the agent. In addition, many works aim to facilitate the emergence of symbol-like representations by using specially designed machine learning architectures. In this thesis, we take a different approach and investigate whether a deep reinforcement learning system that learns a dynamic task would facilitate the formation of high-level neural representations that might be considered as precursors of symbolic representations. For our experiments, we use a simulated robot whose sole task is to learn a policy to stand up under velocity dependent force perturbations. The policy is represented and learned by a neural network that is composed of simulated neurons whose responses are recorded during learning and, later, when testing the policy in different environments. The recorded data are then analyzed to detect formation of high-level abstract representations and their relations to the task dynamics. The results indicate that without even explicit design to promote abstract representations, intriguing neural responses emerge via learning, which may serve as the basis of abstract symbol-like representations.Canlı organizmalar, özellikle memeliler basit planlama gerektiren alet kullanımı ve maniplasyon gibi karmaşık görevleri yerine getirmede ustadırlar. Bu yetenek, merkezi sinir sistemi tarafından, özellikle de aksiyon ve algı için alt seviye sinyal işleme gerçekleştiren nöral devrelerden gelişen beyin tarafından ortaya konmaktadır. Yapay sistemler, temel algılama ve aksiyon alma gerektiren görevler için alt seviye sinyal işlemeyi öğrenebilir. Özellikle pekiştirmeli öğrenme (RL), bir aktörün sensör deneyimi ile bir dizi aksiyon alması ve çevreden dönen "ödülleri" toplayarak belirli bir görevi yerine getirmek için gerekli karar mekanizmasını (policy) oluşturduğu öğrenme şeklidir. Daha gelişmiş hayvanların karmaşık görevleri gerçekleştirme yetenekleri, öğrenme sırasında alt seviye sinir devrelerinin sensör deneyimi döngüsüne dayalı olarak oluşturulan soyutlamalar ile gerçekleştirilir. Benzer bir şekilde bir RL aktörünün öğrenme deneyimi, sembol benzeri yapılar olusturularak, sembol ile planlama gibi karmaşık görevleri yerine getirmesini sağlamak amacı için kullanılabilir. Bu tür yüksek seviyeli yapıların ortaya çıkışı ve bunların çevre ve aktör dinamikleri ile ilişkileri, öğrenme sırasında nöron çıktıları ve muhtemel farklı ortamlarda konuşlandırılmasının analiz edilmesi ile ortaya çıkartılabilir. Literatürde pek çok çalışma bu sembollerin halihazırda var olduğu varsayımı ile başlar ve bu sembolleri sürekli motor sensör çıktıları üzerine oturtur. Diğer bir yaklaşım ise bu sembollerin ortaya çıkmasını sağlamak amacı ile özel olarak dizayn edilmiş sinir ağları kullanmaktır. Bu tez çalışmasında ise dinamik bir görevi öğrenen bir derin pekiştirmeli öğrenme sisteminin üst seviye bir sinirsel yapı oluşturma olasığı incelenmiştir ki bu yapılar sembolik temsillerin öncülleri olarak düşünülebilir. Deneylerimiz için, görevi hız vektörüne bağlı bir bozucu kuvvet altında ayağa kalkmayı öğrenmek olan simüle edilmiş bir robot kullandık. Simule edilmiş nöronlardan oluşan ayağa kalkma mekanizması bir sinir ağı ile öğrenilebilir ve temsil edilebilir. Deneylerimizde böyle bir sinir ağının çıktıları öğrenme sırasında ve sonrasında farklı ortamlarda kaydedilmiştir. Kaydedilen veriler daha sonra üst seviye soyut temsillerin oluşumu ve bunların görev dinamikleri ile ilişkilerini tespit etmek için incelenmiştir. Sonuçlar, soyut yapılar için özel bir tasarım kullanmadan, sembol benzeri gösterim olarak görülebilecek ilgi çekici sinir çıktılarının oluşabileceğini göstermiştir
Event-triggered adaptive handover for centralized hybrid VLCMMW networks
No full textVisible light communication (VLC) builds upon the dual use of LED-based lighting infrastructure for data transmission. It is particularly attractive for user-dense environments and mainly positioned as a complementary wireless access technology to radio-frequency (RF) counterparts. The practical implementation of hybrid lightradio access networks requires the design of efficient vertical and horizontal handover mechanisms to exploit their complementary features. In this paper, we propose an event-triggered adaptive handover mechanism for centralized hybrid VLCmillimeter wave (MMW) systems. Unlike distributed architectures where all the functionalities are deployed in each access point (AP), the centralized systems allow better optimization of network parameters due to awareness of global network view. The proposed handover mechanism takes advantage of this architecture and uses offset-enhanced signal-to-interference plus noise ratio levels of APs as performance metrics. To ensure balanced distribution of user equipment (UE) numbers between two underlying technologies, it auto-tunes the offset variables using rule-set based algorithms based on the number of connected UEs and data rates as inputs. Our results demonstrate that the proposed handover mechanism significantly improves hybrid system performance via adaptive tuning of AP selection parameter while minimizing signaling overhead