Kind Of Architecture Research Articles

The complexity of explaining neural networks through (group) invariants

Ever since the work of Minsky and Papert, it has been thought that neural networks derive their effectiveness by finding representations of the data that are invariant with respect to the task. In other words, the representations eliminate components of the data that vary in a way that is irrelevant. These invariants are naturally expressed with respect to group operations, and thus an understanding of these groups is key to explaining the effectiveness of the neural network. Moreover, a line of work in deep learning has shown that explicit knowledge of group invariants can lead to more effective training results.In this paper, we investigate the difficulty of discovering anything about these implicit invariants. Unfortunately, our main results are negative: we show that a variety of questions around investigating invariant representations are NP-hard, even in approximate settings. Moreover, these results do not depend on the kind of architecture used: in fact, our results follow as soon as the network architecture is powerful enough to be universal. The key idea behind our results is that if we can find the symmetries of a problem then we can solve it.

Impact of fully connected layers on performance of convolutional neural networks for image classification

The Convolutional Neural Networks (CNNs), in domains like computer vision, mostly reduced the need for handcrafted features due to its ability to learn the problem-specific features from the raw input data. However, the selection of dataset-specific CNN architecture, which mostly performed by either experience or expertise is a time-consuming and error-prone process. To automate the process of learning a CNN architecture, this paper attempts at finding the relationship between Fully Connected (FC) layers with some of the characteristics of the datasets. The CNN architectures, and recently datasets also, are categorized as deep, shallow, wide, etc. This paper tries to formalize these terms along with answering the following questions. (i) What is the impact of deeper/shallow architectures on the performance of the CNN w.r.t. FC layers?, (ii) How the deeper/wider datasets influence the performance of CNN w.r.t. FC layers?, and (iii) Which kind of architecture (deeper/shallower) is better suitable for which kind of (deeper/wider) datasets. To address these findings, we have performed experiments with four CNN architectures having different depths. The experiments are conducted by varying the number of FC layers. We used four widely used datasets including CIFAR-10, CIFAR-100, Tiny ImageNet, and CRCHistoPhenotypes to justify our findings in the context of image classification problem. The source code of this work is available at https://github.com/shabbeersh/Impact-of-FC-layers.

Three convolutional neural network models for facial expression recognition in the wild

Facial expression recognition (FER) in the wild is a novel and challenging topic in the field of human emotion perception. Different kinds of convolutional neural network (CNN) approaches have been applied to this topic, but few of them ever considered what kind of architecture was better for the FER research. In this paper, we proposed three novel CNN models with different architectures. The first one is a shallow network, named the Light-CNN, which is a fully convolutional neural network consisting of six depthwise separable residual convolution modules to solve the problem of complex topology and over-fitting. The second one is a dual-branch CNN which extracts traditional LBP features and deep learning features in parallel. The third one is a pre-trained CNN which is designed by transfer learning technique to overcome the shortage of training samples. Extensive evaluations on three popular datasets (public CK+, multi-view BU-3DEF and FER2013 datasets) demonstrated that our models were competitive and representative in the field of FER in the wild research. We achieved significant better results with comparisons to plenty of state-of-the-art approaches. Moreover, we provided discussions on the effectiveness and practicability of CNNs with different feature types and architectures for FER in the wild as well.

An education design architecture for the future Australian doctorate

Doctoral training continues to grow in scale and scope in Australia, but has been subjected to far less design and improvement compared with other facets of higher education. Governments and universities engage in ongoing change which helps respond to opportunities and challenges but also leads to a proliferation of options and approaches. The current research study was seeded and shaped by the ambitious view that despite such refinement the doctorate remains in need of much bolder and deeper design, and particularly design with an education focus. This paper reports outcomes from a four-year national project which sought to articulate a doctoral design architecture. The paper discusses framing contexts and concepts, design and characteristics of the doctoral architecture, then implications for sectoral, institutional and individual practice. It concludes that this kind of architecture can provide a useful guide for growth.

PICTURESQUE POLAROIDS

This paper analyses the Polaroid series Geographical Analogies (2006–2010) by French artist Cyprien Gaillard, one of many recent artworks dealing with the failure of modern architectural utopias. His representations of derelict, decaying locations evoke the tradition of picturesque ruin imagery and the article asks in what way artworks like Gaillard’s update formerly relevant aesthetic categories such as the picturesque. By presenting the Polaroids in vitrines, Gaillard challenges common ideas of what kind of architecture is worthy of portrayal, protection and even monumentalising.Geographical Analogies not only reacquaints us with the picturesque, it also raises questions of moral consequence regarding the appreciation for decay.

Open Living Concept in Barn-House Architecture: Single-Family House Case Studies

’Barn house’ architecture can be presumed as a style or trend in contemporary architecture. The term ‘barn-house’ refers to a house that architecturally refers to barn with a simple, longitudinal form with pitched roof. Originally, barns were converted to houses for practical reasons. Later on, the design aspect was getting more important and any house or any barn were not enough, and so it turned to be a style which can be described as ‘designed barn-house’. The idea appears to be well-recognized: many built houses and house designs represent that style. Also, several online blogs on that subject have been created and published dozens of examples. We may presume that many contemporary architects have either already designed or are aware of designing a barn-house. The reason for that is mostly simplicity of architecture, and lack of details which are common in contemporary architecture. 3 case studies of barn-type house designs have been presented in the article. All of them have been designed by the authors. All the concepts have been designed using synthetic models: physical and virtual. The open space concept can be recognized in all the three case studies: the interior layout seems to be more important than the exterior in that kind of architecture. The authors conclude that some uniform rules can be recognized in barn-type architecture design. However, many concepts have gone far beyond the original barn-house idea. The question is whether that trend can be presumed as distinctive?

Museum in the Countryside: Aesthetics of the Data Centre

As part of OMA's forthcoming exhibition ‘Countryside: Future of the World’, which opens at the Solomon R Guggenheim Museum in New York this autumn, the office is cataloguing the emergence of a new kind of architecture, the data centre. These vast, monolithic facilities are not designed to be inhabited by humans, but through their unimaginable scale and radical starkness they illicit the sensibilities of a new technological sublime. OMA founder Rem Koolhaas explores the ways through which architects might be able to engage this condition through cultural programmes and make it accessible to the people it would otherwise exclude.

Balinese Christian Architecture, 1936-2000

Balinese traditional architecture has grown for hundreds of years. The encounter with Christianity has also taken place long before, but has only made significant changes since the beginning of the 20th century. This paper discusses the process started from the development of Christian village of Blimbingsari and Ambyarsari in West Bali, to a spread-growing period of a new culture and new architecture later known as Balinese Christian Architecture. The purpose of this study is to describe the development of this distinctive activity driven and fascinating architecture, at the span of time where the most important events occurred in the process of inculturation of Hindu-Balinese architecture and western Christianity. This ongoing study is done by a limited field study method on Christian villages in Bali, where the researcher stay, interact and experience the everyday life within the community. The result shows the long life struggle of Balinese Christian Architecture to this modern era. In the end, the experience can be developed to build knowledge regarding inculturation, in creating a new kind of architecture and culture.

Tunable waveguiding in origami phononic structures

A novel design of origami phononic structures with tunable waveguiding features are studied in this research. The structure is formed by attaching cylindrical inclusions on top of origami sheet; in this kind of architecture, folding of the underlying origami sheet can change the periodicity of inclusions. Being periodic in nature, the origami structure exhibits bandgaps features and have the potential to steer wave energy through a path of defects (also known as waveguide) created in the lattice of inclusions. Since the spectral content transmitted through the waveguide depends on the periodicity surrounding it, the folding induced lattice transformation in such origami structures can lead to adaptation in waveguide frequency. In fact, in origami structures that can transform between different Bravais-lattice types, for example between a square and hexagon lattice, the waveguide frequency can be drastically tuned. Such phenomenal adaptation is studied in this research by extracting dispersion diagrams of defective origami structures using the plane wave expansion method. Further, numerical and experimental investigations are performed on finite origami structures with waveguides and the transmission results of different folding configurations demonstrate that waveguide frequency can be significantly altered. Overall, the low effort one-degree-of-freedom (1DOF) folding mechanism combined with the scalable nature of the origami architecture makes the origami phononic structure a novel and effective device for waveguiding applications across a range of frequencies.

Entanglement distribution in star network based on spin chain in diamond

After star network of spins was proposed, generating entanglement directly through spin interactions between distant parties became possible. We propose an architecture which involves coupled spin chains based on nitrogen-vacancy centers and nitrogen defect spins to expand star network. The numerical analysis shows that the maximally achievable entanglement Em exponentially decays with the length of spin chains M and spin noise. The entanglement capability of this configuration under the effect of disorder and spin loss is also studied. Moreover, it is shown that with this kind of architecture, star network of spins is feasible in measurement of magnetic-field gradient.

Quantifying students' perception for deconstruction architecture

Abstract Deconstruction in architecture is like a symbol of liberty. The French philosopher Jacques Derrida started the idea basically in language, and then his idea spread to reach architecture. Deconstruction move produced unique differentiated buildings, where difference was the main idea behind deconstruction. This actually made a deep debate, whether deconstruction was an out of the box philosophy or just a strange architectural composition. The research addressed that this kind of architecture needs complete architectural education to value the philosophy behind it, in addition to highlight how students of architecture in both (juniors level and seniors level), how they perceive deconstruction; an experimental approach was used to find out if the scientific material given in architectural theories about deconstruction may affect the perception levels of the students, these students joined the architectural program at faculty of engineering, Zagazig University, Egypt, and the experiment applied on selected pioneers of deconstruction famous buildings.

The Light Node Communication Framework: A New Way to Communicate Inside Smart Homes.

The Internet of things has profoundly changed the way we imagine information science and architecture, and smart homes are an important part of this domain. Created a decade ago, the few existing prototypes use technologies of the day, forcing designers to create centralized and costly architectures that raise some issues concerning reliability, scalability, and ease of access which cannot be tolerated in the context of assistance. In this paper, we briefly introduce a new kind of architecture where the focus is placed on distribution. More specifically, we respond to the first issue we encountered by proposing a lightweight and portable messaging protocol. After running several tests, we observed a maximized bandwidth, whereby no packets were lost and good encryption was obtained. These results tend to prove that our innovation may be employed in a real context of distribution with small entities.

The Study of the Relationship between Sustainable Architecture and Vernacular Architecture in the North of Iran

This study investigates sustainable architecture and reasons of its usage. Architecture is under influence of a range of socio-economic aspects including but not limited environmental, social, cultural, economical problems and the connection between these issues as well as fashion, religion, climate, and etc. Concerning vernacular architecture and introducing the unseen principles of sustainable architecture in it, are an effective method in order to introduce sustainable architecture, so it will cause renewed focusing on vernacular architecture and through using it converts architectural topics to the modern and international issues. Of course, vernacular architecture was sustainable in its own time and place, so it’s pointless to renouncing it. The north Iranian architecture is a kind of architecture that has maintained its basis of nativism. Principles of sustainable architecture is considered in it, which is studied in this article. This descriptive study’s results let us know the important role of North Iranian vernacular architectures in creating a sustainable approach in architecture.

Modular and Decentralized PCU

This paper presents the studies made on the definition and design of a decentralized and modular electrical architecture of Power Conditioning Unit (PCU).The modular PCU is able to cover a large range of mission demands by adapting the number of power modules (PM) while the electrical interfaces remain the same.A decentralized architecture is proposed where each module is able to control the solar arrays and to manage the batteries. It appears that this kind of architecture becomes feasible thanks to digital circuits and using a communication bus [1].Breadboards are being tested in order to validate the concept.Reliability and robustness aspects are studied and a redundant architecture is tested.

Cuando habitábamos lo elemental. Una mirada crítica sobre la vivienda tradicional en el Chile austral a través de la fotografía del siglo XIX

Photography is, doubtless, an extraordinary tool for the historic study of architecture. This is proven by its capacity to document territories and buildings with ample scales of register. Photography arrived in Chile in the second half of the nineteenth century, coinciding with a process of national construction and territorial consolidation which was being developed in parallel to that of other American nations. Views of cities, streets and buildings, captured by both renowned and unknown photographers, are a testimony of the influence of colonial hispanic architecture on the building traditions of the country. They also document the origins of a new kind of architecture, arising in the so-called ‘colonisation territories’ as a result of an exchange of local and European building traditions, which was swiftly assimilated and expanded. Thanks to the lenses of the nineteenth century photographers today we can take advantage of valuable visual evidence that allows us to better understand the traditional architecture of the centre and South of the country.

The Impact of Opening Sizing on the Airflow Distribution of Double-skin Facade

Double-skin facade (DSF) has been widely used in different kinds of architecture as new building envelopes. The position of the shading louver and opening size of the air layer is the key point to thermal performance of DSF. A temperature test, which is in a DSF building in Wuhan, has been conducted in this paper. The ANSYS software has also been used to simulate the velocity and temperature field in the DSF with different opening size. The results show that the test data keeps well accordance with the simulation. Significant differences exist in the velocity and temperature field in DSF with different shading louver position and opening size. The computing results in summer day indicate that as the opening size varies from 0.1m to 0.6m, temperature and velocity of the air in the layer decrease first, then remain the same. Temperature and velocity distribution becomes more regularly with the louver shading in the middle of the layer or near the outside glass than the region close to the inside glass. The testing data in sunny day keeps well accordance with the simulation result, which means the simulation results have certain reliability.

Graph-Based Simultaneous Localization and Mapping: Computational Complexity Reduction on a Multicore Heterogeneous Architecture

This article deals with the computational complexity issue of graphbased simultaneous localization and mapping (SLAM). SLAM allows a robot that is navigating in an unknown environment to build a map of this environment while simultaneously determining the robot pose on this map. Graph-based SLAM is a smoothing method that uses a graph to represent and solve the SLAM problem. We first propose a graph construction that takes advantage of the incremental and sparse characteristics of graph-based SLAM. This incremental construction is exploited to perform several algorithmic optimizations. Second, we present a study of using a heterogeneous architecture to implement the graph-based SLAM algorithm. Indeed, the emergence of recent heterogeneous embedded architectures should lead to a great advance in the design of embedded systems-based robotics applications. As a result of this study, an algorithm-architecture mapping is proposed for a central processing unit-graphics processing unit (CPU-GPU)-based architecture. The study also investigates how this kind of architecture can speed up graph-based SLAM by offloading some critical compute-intensive tasks of the algorithm on the GPU. Some common data sets are used to compare our implementations to the state of the art.

Mo-Based Transition-Metal-Dichalcogenide Junctionless Field-Effect-Transistors

One of the most promising candidates as substitution for silicon in modern electronics devices are Transition-Metal-Dichalcogenides (TMDs). These are semiconductors in the form of MX2, where M is a transition metal (Mo, W, Hf), and X is a chalcogen (S, Se, Te). The weak Van-Der-Waals force between each layer of which they are formed, allow them to be easily exfoliated by scotch tape technique from their bulk form, as first was shown for graphene.[i] However one of the main limitations of graphene is its zero bandgap and resulting low on/off current ratio, which limits its usefulness for field-effect-transistor (FET) applications. On the other hand, TMDs show a finite direct bandgap between 1-2 eV, predicted by density-functional-theory (DFT) calculations,[ii]which make them suitable for logic devices. Their ultra-thin nature would allow a better electrostatic control for device applications, with respect to the state-of-the-art devices, and better immunity to short-channel effects, earning the attention of the electronics and optoelectronics community. In this work, we report the electrical measurements of a back-gated MoS2 flake doped with Nb, and back-gated undoped MoTe2 mechanical exfoliated from crystals. The flakes were then put on 85 nm thick Si oxide on a highly doped Si handle wafer. Ti/Au (5/45 nm) deposited on top allowed the realization of a back-gate structure, which was analyzed through Scanning Electron Microscopy (SEM) and Transmission Electron Microscopy (TEM). The device was then characterized by voltage and temperature dependence of the current. For the MoS2, the temperature dependency of the device shows semiconductor behavior and, the doping effect by Nb atoms introduces acceptors in the structure, with a concentration 7.44x1019 cm-3 measured by Hall effect. The p-type doping is confirmed by all the electrical measurements, making the structure a junctionless transistor. In addition, other parameters regarding the contact resistance between the metal and MoS2 are extracted thanks to a simple Transfer Length Method structure, showing a promising contact resistivity of 1.07x10-7 Ohm.cm2 and a sheet resistance of 2.36x102 Ω/sq. For the MoTe2 the crystal was nominally undoped, making it easier to electrically fully-deplete, but consequently the on-current, contact resistivity and sheet resistance were worse than in the highly-doped MoS2. In Fig. 1 is a representative SEM image of the contacted MoS2 flake device, with the contact tracks numbered 1 to 4. In Fig. 2. is a representative TEM image of the overall structure of the flake. It is possible to note the undulations of the 10 nm MoS2 flake along the surface of the substrate. Electrical measurements were carried out to confirm the electrical conductivity of the device. Fig 3 shows Ids-Vbgcurve between the first and second pad confirming a p-type field-effect-transistor. In summary, Mo-based TMD flakes were mechanically exfoliated and contacted. The p-type behavior of the MoS2 flake, expected since it is doped with Nb atoms, was confirmed by electrical measurements. The low contact resistance and resistivity data reported, even if extracted with a simple TLM structure, can be related to the high p-type doping of the flake, which, together with other processes widely used to improve metal contacts and obtain clean TMDs surfaces, can be helpful in reaching a better TMD/metal interface. Moreover, the high doping can be the key to the introduction of TMD-based junctionless transistors, in which the good characteristics of TMDs and the uncomplicated process required by this kind of architecture may be combined. [i] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, “Electric Field Effect in Atomically Thin Carbon Films”, Science 306, 666-669 (2004). [ii] O. V. Yazyev and A. Kis, “MoS2 and semicondutors in the flatland”, Materials Today Vol. 18, Number 1 January/February 2015. Figure 1

A Survey on Phase Change Memory-Aware Cache Management

PCM (phase change memory) has been regarded as one of the most promising NVRAM. Due to its excellent features including high access speed, good scalability, byte addressability, and low idle power, PCM is expected to be an alternative to DRAM or flash memory in the future. So far, many different kinds of architecture have been proposed to merge PCM into the memory hierarchy of computer systems. For PCM-based computer systems, considering that PCM has various properties differing with DRAM and conventional storage media, traditional cache management policies based on DRAM and disks have to be re-designed. In this paper, we present a survey on the cache management policies for PCM-based systems that employ PCM as main memory or as secondary storage. Specially, for different PCM-based architectures, we explain how PCM-aware cache management policies exploit PCM’s advantages and overcome PCM’s drawbacks like asymmetric read/write latencies.

Remarks on English Non-Restrictive Relative Clauses

The purpose of this paper is two folds: one is to examine the controversies among the existing research over the syntactic properties of English non-Restrictive Relative Clauses (hereafter, NRRCs) as opposed to Restrictive Relative Clauses (hereafter, RRCs). Secondly, the internal structure of English NRRCs is proposed in order to accommodate the disputes from the previous research such as Quirk, et. al. (1985), Borsley (1992, 1997), Smits (1988), McCawley (1998), Lobeck (2000), Huddleston & Pullum (2008), Vries (2002, 2006), Aoun & Li (2003), Arnold (2004), and Arnold & Borsely (2008), we argue that English NRRCs, unlike RRCs, are XP-adjuncts, (contra Lobeck 2000, Aarts 2001, 2008, among others), arriving at a conclusion that NRRC involves a discontinuous constituent. Furthermore, this Wh-CP in English NRRC enters into the syntactic derivation via Sidewards movement (hereafter, SM) in the sense of Hornstein (2001, 2009) in order to eventually form a single-rooted tree. The kind of architecture we advocate for English NRRC in this study is on a par with what Newson (2006) has argued for: English RRC is an X’-Adjunct and NRRC is an XP-Adjunct. Most of the troublesome properties that have long been discussed in literature receive a new analysis under the current study. For example, the obviation of WCO effect or ‘weakest crossover’ (Lasnik & Stowell 1991) observed in English NRRC, unlike in RRCs (Authier & Reed 2005), can be straightforwardly accounted for, if the two distinct structures are to be adopted. The theoretical implication of this proposal is that there are two types of adjuncts, one is X’ adjunct and the other is XP-adjunct, contra Hornstein (2009).