Choice Of Representation Research Articles

3D deep convolutional neural networks for amino acid environment similarity analysis

BackgroundCentral to protein biology is the understanding of how structural elements give rise to observed function. The surfeit of protein structural data enables development of computational methods to systematically derive rules governing structural-functional relationships. However, performance of these methods depends critically on the choice of protein structural representation. Most current methods rely on features that are manually selected based on knowledge about protein structures. These are often general-purpose but not optimized for the specific application of interest.In this paper, we present a general framework that applies 3D convolutional neural network (3DCNN) technology to structure-based protein analysis. The framework automatically extracts task-specific features from the raw atom distribution, driven by supervised labels. As a pilot study, we use our network to analyze local protein microenvironments surrounding the 20 amino acids, and predict the amino acids most compatible with environments within a protein structure. To further validate the power of our method, we construct two amino acid substitution matrices from the prediction statistics and use them to predict effects of mutations in T4 lysozyme structures.ResultsOur deep 3DCNN achieves a two-fold increase in prediction accuracy compared to models that employ conventional hand-engineered features and successfully recapitulates known information about similar and different microenvironments. Models built from our predictions and substitution matrices achieve an 85% accuracy predicting outcomes of the T4 lysozyme mutation variants. Our substitution matrices contain rich information relevant to mutation analysis compared to well-established substitution matrices. Finally, we present a visualization method to inspect the individual contributions of each atom to the classification decisions.ConclusionsEnd-to-end trained deep learning networks consistently outperform methods using hand-engineered features, suggesting that the 3DCNN framework is well suited for analysis of protein microenvironments and may be useful for other protein structural analyses.

Starting From … Now and the web series to television crossover: an online revolution?

Over the past two decades, television has moved rapidly into an era marked by increased choice, flexibility and greater diversity of representation. Coinciding with this is a rise in new storytelling forms, such as web series, which have gained traction online and via traditional means. These developments have also manifested the phenomenon of the web series to television crossover. This article looks to Starting From … Now (2014–2016), a lesbian-themed web series that was broadcast on SBS2 in 2016 and, subsequently, described as a ‘revolution’ in Australian television. Examining the industrial context and critical reception of the series, this article assesses the validity of this claim. This research elucidates the place of web series in Australia’s shifting television landscape and highlights how Starting From … Now confronts the lack of diversity in Australian TV drama.

Semicrossed products of operator algebras by semigroups

We examine the semicrossed products of a semigroup action by $*$-endomorphisms on a C*-algebra, or more generally of an action on an arbitrary operator algebra by completely contractive endomorphisms. The choice of allowable representations affects the corresponding universal algebra. We seek quite general conditions which will allow us to show that the C*-envelope of the semicrossed product is (a full corner of) a crossed product of an auxiliary C*-algebra by a group action. Our analysis concerns a case-by-case dilation theory on covariant pairs. In the process we determine the C*-envelope for various semicrossed products of (possibly nonselfadjoint) operator algebras by spanning cones and lattice-ordered abelian semigroups. In particular, we show that the C*-envelope of the semicrossed product of C*-dynamical systems by doubly commuting representations of $\mathbb{Z}^n_+$ (by generally non-injective endomorphisms) is the full corner of a C*-crossed product. In consequence we connect the ideal structure of C*-covers to properties of the actions. In particular, when the system is classical, we show that the C*-envelope is simple if and only if the action is injective and minimal. The dilation methods that we use may be applied to non-abelian semigroups. We identify the C*-envelope for actions of the free semigroup $\mathbb{F}_+^n$ by automorphisms in a concrete way, and for injective systems in a more abstract manner. We also deal with C*-dynamical systems over Ore semigroups when the appropriate covariance relation is considered.

Exploiting vector instructions with generalized stream fusion

Ideally, a program written as a composition of concise, self-contained components should perform as well as the equivalent hand-written version where the functionality of what was many components has been manually combined into a monolithic implementation. That is, programmers should not have to sacrifice code clarity or good software engineering practices to obtain performance---we want compositionality without a performance penalty. This work shows how to attain this goal for high-level Haskell in the domain of sequence-processing functions, which includes applications such as array processing. Prior work on stream fusion shows how to automatically transform some high-level sequence-processing functions into efficient implementations. It has been used to great effect in Haskell libraries for manipulating byte arrays, Unicode text, and unboxed vectors. However some operations, like vector append, do not perform well within the stream fusion framework. Others, like SIMD computation using the SSE and AVX instructions available on modern x86 chips, do not seem to fit in the stream fusion framework at all. We describe generalized stream fusion, which solves these issues through a careful choice of stream representation. Benchmarks show that high-level Haskell code written using our compiler and libraries can produce code that is faster than both compiler- and hand-vectorized C.

Earthquake‐induced loss assessment of steel frame buildings with special moment frames designed in highly seismic regions

SummaryThis paper discusses an analytical study that quantifies the expected earthquake‐induced losses in typical office steel frame buildings designed with perimeter special moment frames in highly seismic regions. It is shown that for seismic events associated with low probabilities of occurrence, losses due to demolition and collapse may be significantly overestimated when the expected loss computations are based on analytical models that ignore the composite beam effects and the interior gravity framing system of a steel frame building. For frequently occurring seismic events building losses are dominated by non‐structural content repairs. In this case, the choice of the analytical model representation of the steel frame building becomes less important. Losses due to demolition and collapse in steel frame buildings with special moment frames designed with strong‐column/weak‐beam ratio larger than 2.0 are reduced by a factor of two compared with those in the same frames designed with a strong‐column/weak‐beam ratio larger than 1.0 as recommended in ANSI/AISC‐341‐10. The expected annual losses (EALs) of steel frame buildings with SMFs vary from 0.38% to 0.74% over the building life expectancy. The EALs are dominated by repairs of acceleration‐sensitive non‐structural content followed by repairs of drift‐sensitive non‐structural components. It is found that the effect of strong‐column/weak‐beam ratio on EALs is negligible. This is not the case when the present value of life‐cycle costs is selected as a loss‐metric. It is advisable to employ a combination of loss‐metrics to assess the earthquake‐induced losses in steel frame buildings with special moment frames depending on the seismic performance level of interest. Copyright © 2017 John Wiley & Sons, Ltd.

What Makes an Effective Representation of Information: A Formal Account of Observational Advantages

In order to effectively communicate information, the choice of representation is important. Ideally, a chosen representation will aid readers in making desired inferences. In this paper, we develop the theory of observation: what it means for one statement to be observable from another. Using observability, we give a formal characterization of the observational advantages of one representation of information over another. By considering observational advantages, people will be able to make better informed choices of representations of information. To demonstrate the benefit of observation and observational advantages, we apply these concepts to set theory and Euler diagrams. In particular, we can show that Euler diagrams have significant observational advantages over set theory. This formally justifies Larkin and Simon’s claim that “a diagram is (sometimes) worth ten thousand words”.

A study of metrics of distance and correlation between ranked lists for compositionality detection

Compositionality in language refers to how much the meaning of some phrase can be decomposed into the meaning of its constituents and the way these constituents are combined. Based on the premise that substitution by synonyms is meaning-preserving, compositionality can be approximated as the semantic similarity between a phrase and a version of that phrase where words have been replaced by their synonyms. Different ways of representing such phrases exist (e.g., vectors (Kiela and Clark, 2013) or language models (Lioma, Simonsen, Larsen, and Hansen, 2015)), and the choice of representation affects the measurement of semantic similarity.We propose a new compositionality detection method that represents phrases as ranked lists of term weights. Our method approximates the semantic similarity between two ranked list representations using a range of well-known distance and correlation metrics. In contrast to most state-of-the-art approaches in compositionality detection, our method is completely unsupervised. Experiments with a publicly available dataset of 1048 human-annotated phrases shows that, compared to strong supervised baselines, our approach provides superior measurement of compositionality using any of the distance and correlation metrics considered.

How to make ‘more’ better? Principles for effective use of multiple representations to enhance students’ learning about fractions

To make complex mathematics concepts accessible to students, teachers often rely on visual representations. Because no single representation can depict all aspects of a mathematics concept, instruction typically uses multiple representations. Much research shows that multiple representations can have immense benefits for students’ learning. However, some re-search also cautions that multiple representations may fail to enhance students’ learning if they are not used in the “right” way. For example, unless students can (1) properly interpret each individual representation and (2) make connections among multiple representations and the information they intend to convey, the use of multiple representations may actually confuse students rather than aid their learning. In this article, we review research-based principles for how to use multiple representations effectively so that they enhance student learning. Using fractions as an illustrative domain, we discuss how the choice of visual representation may affect student learning based on the conceptual aspects of the to-be-learned content emphasized by the representation. Next, we describe ways to help students interpret individual representations and to make connections among them. We illustrate these arguments with our own empirical research on fractions learning. We conclude by laying out open questions that future research should address.

Learning Legible Motion from Human–Robot Interactions

In collaborative tasks, displaying legible behavior enables other members of the team to anticipate intentions and to thus coordinate their actions accordingly. Behavior is therefore considered to be legible when an observer is able to quickly and correctly infer the intention of the agent generating the behavior. In previous work, legible robot behavior has been generated by using model-based methods to optimize task-specific models of legibility. In our work, we rather use model-free reinforcement learning with a generic, task-independent cost function. In the context of experiments involving a joint task between (thirty) human subjects and a humanoid robot, we show that: (1) legible behavior arises when rewarding the efficiency of joint task completion during human–robot interactions (2) behavior that has been optimized for one subject is also more legible for other subjects (3) the universal legibility of behavior is influenced by the choice of the policy representation.

Different patterns of university students\u2019 integration of lecture podcasts, learning materials, and lecture attendance in a psychology course

Lecture podcasts are considered an efficient means for passing on learning contents to students, most notably in lectures with large numbers of students. Here, the lecturer’s presentation, combined with lecture slides, is recorded and broadcasted in video form. The present study investigates how students organize learning when they have the choice of different representations of content: face-to-face lectures, lecture recordings as video podcasts, and additional text material. Latent class analysis identified three groups with different patterns of integration of these representations of content: (1) students who focused on podcasts; (2) students who made little use of the different representations of content; (3) students who occasionally made use of the different learning opportunities. Students in group 1 with a focus on podcasts achieved best. They devoted more time to learning and made diligent use of a variety of learning strategies such as note-taking, generating summaries, or rehearsing with the podcasts. The results suggest providing different representations of content to take into account differences in learners’ preferences and abilities. They speak in favor of podcasts, especially in lectures with a large audience. However, the results also show a group of learners who make minimal use of the various representations of content. Instructional measures should be taken to engage and support these students.

The Delaware Trap: An Empirical Study of Incorporation Decisions

One of the most enduring debates in corporate law centers on why Delaware has become the dominant state in the market for corporate charters. Traditionally, two perspectives dominated the debate, the “race to the top” perspective that sees competition among states as driving legal rules toward efficiency and the “race to the bottom” perspective that sees competition among states as driving legal rules toward the interests of corporate managers. The two dominant perspectives have struggled to explain why approximately half of companies incorporate in Delaware, while the other half incorporate in their home states. Whether the choices are attributable to the quality of state law or to characteristics of the companies themselves or both has given rise to a large, but inconclusive empirical literature.This Article uses a large dataset of corporate financings to shed new light on this mystery and uncovers strong evidence that some of the strongest factors in incorporation choice are factors unrelated to either the quality of state law or the characteristics of individual companies. Instead, the data strongly suggest that demographic markers of sophistication, such as choice of law firm and headquarters location, predict the jurisdictional choice about as well as state law or the business attributes of companies. Companies with more demographic markers of sophistication tend to choose Delaware incorporation, and companies with fewer demographic markers of sophistication tend to choose home-state incorporation. The finding persists even when other attributes of the company are controlled for, such as its industry classification, the amount of money raised, or whether the company is public or private. Indeed, the sophistication factors arguably predict Delaware incorporation as well or better than any factors documented in the vast literature on state competition for corporate charters.The findings have important implications for the state “race-to-the-top” debate in corporate law. At a minimum the results in this Article make it clear that the choice of legal representation is an important missing variable in models of incorporation decisions. The fact that the choice of law firms drives the jurisdictional choice has far broader implications. If law firms drive the jurisdictional choice they may steer companies toward states that serve the law firms’ own interests without regard to the quality of legal rules or the needs of the client. When the state chosen is Delaware, as it often is, there are few alternative jurisdictions that shareholders and managers can agree on. As a result, companies inadvertently fall into a “governance trap” from which reincorporation out of state is nearly impossible. This interpretation would suggest that Delaware’s carefully calibrated positioning in the charter market has largely eliminated meaningful competition among the states for the quality of corporate law.

Non-Hermitian surface hopping.

We present a generalized non-Hermitian equation of motion (nH-EOM) to go beyond standard trajectory surface hopping dynamics. The derivation is based on the Born-Huang expansion of the total wave function and the polar representation of the nuclear factor. The nH-EOM contains two additional terms, a skew symmetry term iΓ with dissipation operator Γ to account for decoherence, and a kinetic-energy renormalization term to account for phase shifts, without destroying the invariance to the choice of representation. Numerically, the nH-EOM can still be solved efficiently using a semiclassical approximation in the framework of Tully's fewest-switches surface hopping (FSSH) algorithm. Applications to model Hamiltonians demonstrate improved performance over the standard FSSH approach, through comparison to exact quantum results.

Identifying the Influence Factors on the Intention to Vote among Youths in Rural Areas

Ever since the new millennium, registration rates and voter turnouts for the general elections among Malaysian youths between the age range of 21 and 22 years have been low and dismaying, particularly so in non-urban locations. With the number of youths forming a significant percentage of Malaysia’s population, targeting eligible youths for the various political parties is imperative in winning the elections. Furthermore, youth participation and engagement in national politics have been impactful in the direction and development of Malaysia’s government and administration. Therefore, the purpose of this study is to identify the influence factors on the intention to vote in the next general election among youths in rural areas. The study was conducted using a quantitative design, crosssectional survey. The unit of analysis was identified through stratified sampling technique. From a sample size of 400 (n=400), the returned rate was 370 or 92.5%. The findings revealed that there was a significant relationship between the factors of youth’s to vote in the next general election based on four variables; awareness, persuading method, spatial representation, and political vote choice. The regression analysis ranked awareness as the most influential factor in the youths’ intentions to vote. The study also found that the moderating role of gender is the most significant factor among the four influence factors. The findings suggest that the rural youths’ intentions to vote in the next general election would be intensified if regular promotions are done to enhance awareness. Therefore, the relevant stakeholders should take note of this when campaigning for their party. Aside from that, the findings provide additional literature to the body of knowledge on the political inclination of the younger generation.

Conflicting accounts of λ-definability

A function on some domain is λ-definable if the corresponding function of λ-terms is so definable. However, the correspondence is parametrized by a representation of the domain. Often there is a natural choice of representation, but when the domain consists of λ-terms then they can be represented by either themselves or by the Church numeral of their Gödel number. This choice determines whether or not all computable functions are λ-definable.

Parametric Surface Representation with Bump Image for Dense 3D Modeling Using an RBG-D Camera

When constructing a dense 3D model of an indoor static scene from a sequence of RGB-D images, the choice of the 3D representation (e.g. 3D mesh, cloud of points or implicit function) is of crucial ...

Party Splits, Not Progressives

The choice of proportional representation (PR) is rarely included in work on American local politics. Yet we have long known that 24 cities adopted the single transferable vote form of PR from 1915 to 1948. Breaking with a machine–reform dichotomy that dominates the PR historiography, I investigate two partisan hypotheses about PR’s origins. One concerns the emergence of third parties. A second involves splits in ruling parties. In at least 15 cases, PR choice involved an alliance of convenience between ruling-party defectors and local minority parties. Evidence includes narratives on the partisanship of elite PR backers, comparison of case history and precinct-level referendum outcomes for three similar cities, and aggregate data on big-city charter change referenda from 1900 to 1950. New in this article is comparison of PR adopters with non-adopters. Party splits in places with sizable out-parties emerge as a distinctly American path to proportional electoral rules.1

Communication: Understanding molecular representations in machine learning: The role of uniqueness and target similarity.

The predictive accuracy of Machine Learning (ML) models of molecular properties depends on the choice of the molecular representation. Inspired by the postulates of quantum mechanics, we introduce a hierarchy of representations which meet uniqueness and target similarity criteria. To systematically control target similarity, we simply rely on interatomic many body expansions, as implemented in universal force-fields, including Bonding, Angular (BA), and higher order terms. Addition of higher order contributions systematically increases similarity to the true potential energy and predictive accuracy of the resulting ML models. We report numerical evidence for the performance of BAML models trained on molecular properties pre-calculated at electron-correlated and density functional theory level of theory for thousands of small organic molecules. Properties studied include enthalpies and free energies of atomization, heat capacity, zero-point vibrational energies, dipole-moment, polarizability, HOMO/LUMO energies and gap, ionization potential, electron affinity, and electronic excitations. After training, BAML predicts energies or electronic properties of out-of-sample molecules with unprecedented accuracy and speed.

Newman–Janis Ansatz in conformastatic spacetimes

The Newman–Janis Ansatz was used first to obtain the stationary Kerr metric from the static Schwarzschild metric. Many works have been devoted to investigate the physical significance of this Ansatz, but no definite answer has been given so far. We show that this Ansatz can be applied in general to conformastatic vacuum metrics, and leads to stationary generalizations which, however, do not preserve the conformal symmetry. We investigate also the particular case when the seed solution is given by the Schwarzschild spacetime and show that the resulting rotating configuration does not correspond to a vacuum solution, even in the limiting case of slow rotation. In fact, it describes in general a relativistic fluid with anisotropic pressure and heat flux. This implies that the Newman–Janis Ansatz strongly depends on the choice of representation for the seed solution. We interpret this result as a further indication of its applicability limitations.

Centre-valued Index for Toeplitz Operators with Noncommuting Symbols

AbstractWe formulate and prove a “winding number” index theorem for certain “Toeplitz” operators in the same spirit as Gohberg–Krein, Lesch and others. The “number” is replaced by a self-adjoint operator in a subalgebraZ⊆Z(A) of a unital C*-algebra,A. We assume a faithfulZ-valued traceτonAleft invariant under an actionα:R → Aut(A) leavingZpointwise fixed. Ifδis the infinitesimal generator ofαanduis invertible in dom(δ), then the “winding operator” ofuis. By a careful choice of representations we extend (A, Z, τ, α) to a von Neumann settingwhereand. Then, the von Neumann crossed product, and there is a faithful, dual-trace on. IfPis the projection incorresponding to the non-negative spectrum of the generator ofRinsideandis the embedding, then we defineand show it is Fredholm in an appropriate sense and the-valued index ofTuis the negative of the winding operator. In outline the proof follows that of the scalar case done previously by the authors. The main difficulty is making sense of the constructions with the scalars replaced byin the von Neumann setting. The construction of the dual-trace onrequires the nontrivial development of a-Hilbert algebra theory. We show that certain of these Fredholm operators fiber as a “section” of Fredholm operators with scalar-valued index and the centre-valued index fibers as a section of the scalar-valued indices.

The influence of sorption hysteresis on diffusion coefficients represented with different moisture potentials

This paper continues the discussion in a previous paper [Saeidpour and Wadsö doi:10.1016/j.cemconres.2016.02.003] where a clear effect of sorption hysteresis was seen in moisture diffusion coefficients for cement-based materials. Coefficients with vapor content as potential (Dv) had different values depending on whether the sample was in absorption or in desorption. In this paper we recalculate these Dv coefficients into coefficients with other potentials and draw them as a function of different potentials. The aim is to see if the effect of hysteresis disappears with an optimal choice of representation. The best choice was to draw Dv is as a function of the degree of saturation (concentration, moisture content).