Compositional Task Research Articles

Compositional pretraining improves computational efficiency and matches animal behavior on complex tasks.

1Recurrent neural networks (RNN) are ubiquitously used in neuroscience to capture both neural dynamics and behaviors of living systems. However, when it comes to complex cognitive tasks, training RNNs with traditional methods can prove difficult and fall short of capturing crucial aspects of animal behavior. Here we propose a principled approach for identifying and incorporating compositional tasks as part of RNN training. Taking as target a temporal wagering task previously studied in rats, we design a pretraining curriculum of simpler cognitive tasks that reflect relevant sub-computations. We show that this pretraining substantially improves learning efficacy and is critical for RNNs to adopt similar strategies as rats, including long-timescale inference of latent states, which conventional pretraining approaches fail to capture. Mechanistically, our pretraining supports the development of slow dynamical systems features needed for implementing both inference and value-based decision making. Overall, our approach is an important step for endowing RNNs with relevant inductive biases, which is important when modeling complex behaviors that rely on multiple cognitive computations.

Are All Pictures Worth 1,000 Words? An Investigation of Fit Between Graph Type and Performance on Accounting Data Analytics Tasks

ABSTRACT Despite the increasing prevalence of the use of data analytics and visualizations in accounting settings, little is known about the relative effectiveness of different visualization formats for tasks such as those involving comparisons over time and those involving evaluations of whole-part compositions. In this study, we examine if there is a difference in decision efficiency and effectiveness between bar and line graphs for a comparison task and between pie and stacked column graphs for a compositional task. We also examine if the component complexity of graphs affects performance. We find that line graphs result in significantly greater decision sensitivity (ability to identify true positives) than bar graphs on comparison tasks regardless of the degree of component complexity. We also find that bar graphs result in significantly greater decision specificity (ability to identify true negatives) and overall accuracy.

Learning Compositional Tasks from Language Instructions

Learning Compositional Tasks from Language Instructions

MER: Modular Element Randomization for robust generalizable policy in deep reinforcement learning

Improving the generalization ability of reinforcement learning (RL) agents is an open and challenging problem and has gradually received attention in recent years. Previous work attempts to partially solve this problem by recombining several learned policies to achieve compositional generalization. However, these methods are usually limited to tasks that are composed of fixed subtasks with no task-agnostic environment change. In this paper, we propose a more flexible method termed MER to learn a compositionally generalizable policy that depends on task-dependent invariant elements among tasks, instead of depending on fixed subtasks. As a result, our learned policy can overcome the task-agnostic change in the environment and generalize to more different compositional tasks. Theoretical analysis and experimental results show that our method outperforms traditional methods and exhibits superior generalization ability in unseen new tasks.

Creativity in Education from the Teacher’s Perspective and Its Possible Implications for Music Creative Work with Children Aged 5–6 in Formal Education in the United Kingdom

This study is a qualitative research project focused on teachers’ views of their influences on children’s (aged 5–6) musical creativity within the context of formal education in the United Kingdom. The study offers insight into teachers’ efforts to develop creativity directly and indirectly and explores other factors which connect to creativity, such as institutional context, teacher training, and musical and pedagogical skills. Participants (seven music teachers work- ing in state and private schools in York, UK, with children aged 5–6) were recruited through snowball sampling and participated in one-to-one semi-structured interviews. The findings suggest that teachers implemented age-appropriate short improvisational and compositional tasks. However, free access to musical instruments was supported in nurseries rather than in Year 1 of formal school education. Interviewees considered themselves creative, musical, and musically creative teachers, although only some of them confirmed that their teacher training helped them develop children’s musical creativity.

Stylization and transformation of an image in the aspect of perception of an artistic image

The article discusses such aspects of the topic as the concept of stylization and transformation, the relationship of stylization with the concept of style in fine art, as well as its impact on the perception of an artistic image. The question of visual and emotional perception is touched upon, which are directly influenced by a stylized artistic image. The subject of the article is the analysis of stylization methods and techniques that work on the perception of an artistic image. The aim of the research is to comprehend the ways and techniques of stylization of the depicted object, through artistic means of expression, which in the perceived content are able to create a more emotional and expressive artistic image. Depending on the degree of transformation of the surrounding reality, the author identified such stylization methods as the iconic method, the geometric method, symbol-based stylization and the abstract method of stylization. Based on a specific artistic material, the analysis made it possible to identify the peculiarities of the perception of an artistic image with each method of stylization. Stylization techniques that work on the perception of the viewer, due to the achievement of greater expressiveness of the artistic image, are determined. The author emphasizes the need to use stylization and transformation to endow an artistic image with increased expression and originality, which in the structure of an artistic work are the main means of reflecting an artistic idea and an integral part in the performance of the set compositional tasks.

A Modular Vision Language Navigation and Manipulation Framework for Long Horizon Compositional Tasks in Indoor Environment.

In this paper we propose a new framework—MoViLan (Modular Vision and Language) for execution of visually grounded natural language instructions for day to day indoor household tasks. While several data-driven, end-to-end learning frameworks have been proposed for targeted navigation tasks based on the vision and language modalities, performance on recent benchmark data sets revealed the gap in developing comprehensive techniques for long horizon, compositional tasks (involving manipulation and navigation) with diverse object categories, realistic instructions and visual scenarios with non reversible state changes. We propose a modular approach to deal with the combined navigation and object interaction problem without the need for strictly aligned vision and language training data (e.g., in the form of expert demonstrated trajectories). Such an approach is a significant departure from the traditional end-to-end techniques in this space and allows for a more tractable training process with separate vision and language data sets. Specifically, we propose a novel geometry-aware mapping technique for cluttered indoor environments, and a language understanding model generalized for household instruction following. We demonstrate a significant increase in success rates for long horizon, compositional tasks over recent works on the recently released benchmark data set -ALFRED.

Learning Parameterized Task Structure for Generalization to Unseen Entities

Real world tasks are hierarchical and compositional. Tasks can be composed of multiple subtasks (or sub-goals) that are dependent on each other. These subtasks are defined in terms of entities (e.g., "apple", "pear") that can be recombined to form new subtasks (e.g., "pickup apple", and "pickup pear"). To solve these tasks efficiently, an agent must infer subtask dependencies (e.g. an agent must execute "pickup apple" before "place apple in pot"), and generalize the inferred dependencies to new subtasks (e.g. "place apple in pot" is similar to "place apple in pan"). Moreover, an agent may also need to solve unseen tasks, which can involve unseen entities. To this end, we formulate parameterized subtask graph inference (PSGI), a method for modeling subtask dependencies using first-order logic with factored entities. To facilitate this, we learn parameter attributes in a zero-shot manner, which are used as quantifiers (e.g. is_pickable(X)) for the factored subtask graph. We show this approach accurately learns the latent structure on hierarchical and compositional tasks more efficiently than prior work, and show PSGI can generalize by modelling structure on subtasks unseen during adaptation.

COMPOSITION IN THE LANDSCAPE: ART AND PEDAGOGICAL ASPECT

The author emphasizes the importance of studying the landscape as a genre of fine arts by students. The laws and techniques of composition are the basis, the basis of writing works of all genres of painting that is why compositional competencies are the main place in the structure of art education. This is especially true of landscape composition.Creating a landscape involves drawing in the open air, outdoors. The work should start with finding the best location – landscaping locations. It is well known in an art practice that any composition begins with sketches. The same goes for landscape composition. It is emphasized that in the course of future teachers' of fine arts professional training should be acquainted with the basic laws of the construction of landscape composition on the plane, the means of applying the method of golden intersection in landscapes, methods of reaching the depth of space in the image of the surrounding nature in different compositional formats and compositional generalization; step-by-step work from compositional searches to detailing and completion of a complete artistic landscape image. Art educators argue that all the details in the landscape should be well connected and helped to express the overall idea of the composition. To do this, students need to focus on more area characteristic features, select the most typical landscape objects. It is revealed that the landscape is not just a copy of nature from nature, but also the expression of the artist's attitude to nature, the transfer of the artist's own mood, their experiences and emotions. The author emphasizes the importance of studying the соmpositional techniques in the landscape should be applied not formally, but in order to best express the main idea of the artist. The author agrees with the opinion of artist-educators, who argue that students should be given the opportunity to independently set certain compositional tasks and perform them independently in the creation of a landscape.

*-CFQ: Analyzing the Scalability of Machine Learning on a Compositional Task

We present *-CFQ ("star-CFQ"): a suite of large-scale datasets of varying scope based on the CFQ semantic parsing benchmark, designed for principled investigation of the scalability of machine learning systems in a realistic compositional task setting. Using this suite, we conduct a series of experiments investigating the ability of Transformers to benefit from increased training data size under conditions of fixed computational cost. We show that compositional generalization remains a challenge at all training sizes, and we show that increasing the scope of natural language leads to consistently higher error rates, which are only partially offset by increased training data. We further show that while additional training data from a related domain improves the accuracy in data-starved situations, this improvement is limited and diminishes as the distance from the related domain to the target domain increases.

Multi-task Hierarchical Reinforcement Learning for Compositional Tasks

Multi-task Hierarchical Reinforcement Learning for Compositional Tasks

Evolutionary music: applying evolutionary computation to the art of creating music

We present a review of the application of genetic programming (GP) and other variations of evolutionary computation (EC) to the creative art of music composition. Throughout the development of EC methods, since the early 1990s, a small number of researchers have considered aesthetic problems such as the act of composing music alongside other more traditional problem domains. Over the years, interest in these aesthetic or artistic domains has grown significantly. We review the implementation of GP and EC for music composition in terms of the compositional task undertaken, the algorithm used, the representation of the individuals and the fitness measure employed. In these aesthetic studies we note that there are more variations or generalisations in the algorithmic implementation in comparison to traditional GP experiments; even if GP is not explicitly stated, many studies use representations that are distinctly GP-like. We determine that there is no single compositional challenge and no single best evolutionary method with which to approach the act of music composition. We consider autonomous composition as a computationally creative act and investigate the suitability of EC methods to the search for creativity. We conclude that the exploratory nature of evolutionary methods are highly appropriate for a wide variety of compositional tasks and propose that the development and study of GP and EC methods on creative tasks such as music composition should be encouraged.

The Voice

The Voice R. M. Berry (bio) Stanley Cavell and the Potencies of the Voice Adam Gonya Bloomsbury Academic www.bloomsbury.com 224 Pages; Cloth, $84.00 In Stanley Cavell and the Potencies of the Voice, Adam Gonya offers a new way of understanding Stanley Cavell's philosophy, situating it relative to two "pictures" of the human voice. According to the first, a speaker or writer knows something and selects the right words to communicate it. According to the second, a speaker or writer surrenders to a mood and words simply come. Gonya calls the former the picture of the voice's "assertive potency" and the latter the picture of its "receptive potency," and he takes Schopenhauer and Nietzsche to exemplify each. While the metaphysical alienation imagined by both Arthur Schopenhauer and Friedrich Nietzsche seems without obvious parallel in Cavell's work, what interests Gonya is the role of language in the two philosophers' responses to it. In Schopenhauer's philosophy, a poet escapes metaphysical confinement by perceiving "the particular thing," not as an example under a concept, but as "the [transcendental] Idea of its species." Only after having attained this "pure knowledge" does he or she face the problem of how to express it. The compositional task becomes one of using words to qualify and limit other words, "as one dull knife can sharpen another," producing a distinct representation freed of all abstraction. Theoretically, it could be rendered in prose as well as poetry, since the material qualities of language—rhythm, alliteration, accent, syntax, rhyme—add nothing to it. Its meaning is translatable into other languages without loss. While this account of satisfactory expression bears some similarity to Logical Positivism's picture of ideal reference, it diverges sharply from Nietzsche's account of Dionysian poetry. For Nietzsche, the Dionysian utterance has no prose equivalent or adequate translation. Instead, it originates in a state of intoxication, a "musical mood" that is fundamentally alien to words and concepts. Rather than sharpening meaning, the poet's words only expose their meaning's inadequacy, as though pursuing an impossible absoluteness of differentiation and "straining [language] to its limits to imitate music." While it is not clear that this pursuit gives verbal sounds priority over sense, it is clear that Dionysian poetry, like music, is without articulable content. The poet is a medium or channel merely, with no knowledge of his or her own to impart. As Gonya put it, "The Dionysian utterance … is its singularity." Cavell's importance for Gonya is that his ordinary language procedure, asking what we say when, represents an overlap of these two pictures. Gonya interprets Cavell's ambition to "bring the human voice back into philosophy" as "a therapeutic broadening of what we have called the first potency to include the resources of the second," a broadening that exposes the "sometimes tragic" consequences of philosophy's restriction to justified assertion alone. To bring out these consequences, Gonya introduces three figures—"Jungle Man," "Skeptical Man," and "Common Man"—which he believes make concrete the "zones of expressive strain or miscarriage or even failure" that Cavell addresses. In discussing each, Gonya detects a struggle in Cavell's writing between the imperative to make oneself intelligible and the imperative to avoid falsifying oneself, and if, as Gonya suggests, Cavell proceeds more by receptiveness to what we say when than by testing it against antecedent knowledge, the metaphysical question of the origin of what we say when becomes hard to ignore. It is hard not to feel that Jungle Man is less germane to Cavell's philosophy than Gonya's other figures, since Cavell never treats a lack of training in or exposure to language as a philosophical problem. In that sense, Jungle Man may represent a zone of expressive failure less than a philosophical fantasy of such failure, one that Wittgenstein's therapies would try to make "completely disappear." The fantasy is of someone native to our community who because of some extraordinary experience—Jungle Man wanders off as a child and then, after miraculously surviving in the wild for fifteen years, returns home—finds our ordinary expressive practices unnatural. "Of what had he been deprived that now...

Light factor in the city’s architectural environment

Lighting design is the area regulated today only indirectly, proceeding from requirements of visibility of objects, mainly, criteria of quantity of light on road coverings and facades of the lit objects and uneven distribution on them. There is an underestimation of its role and potential in creating an artistic image of the city in the design of its architectural environment, counting on the visual perception of it only in the daytime. Purpose of research. To identify and formulate problems and prospects for the development of lighting design, to develop recommendations for the light solution of the urban environment. To consider the light environment as a system, to determine the main levels of architectural and compositional tasks of its formation. Method of research. With the help of the analysis of the light environment of the city of Belgorod the problems of its formation are revealed. The main disadvantage of the formation of the urban light environment is the lack of unity in the solution of lighting different areas of the city, the prevalence of light advertising over architectural objects, destroying the stylistic unity of the urban environment. The comparison of the solution of such problems by the Metropolitan administration is carried out. The importance of light as a factor in the formation of the originality of the city is noted. Research result. The development of the concept of urban lighting is proposed, its basic principles are highlighted, ways of solving the problem of organizing night lighting of the urban environment are identified. The potential of lighting design and lighting technologies in modern architecture is only half, and in Russian projects – hardly a quarter. We are in the beginning, but a growing number of interesting projects gives reason to believe that we are moving in the right direction and with the right approach, using the expertise already available, by purchasing and accumulating your own will inevitably achieve the development of their professional achievements in the field of lighting design and the organization of the light environment of the city. Currently, the problem of competent and qualified approach to urban lighting requires the full attention of architects-urban planners and architectural and design organizations.

A mixed-methods study to identify effective practices in the teaching of writing in a digital learning environment in low income schools

This paper reports on the teaching practices identified as effective for students' writing progress in a digital learning environment. The study is situated within a design-based research partnership between researchers and a group of urban schools serving culturally diverse students from low income communities who have implemented a digital pedagogy innovation which includes student device ownership, wireless access and a shared pedagogical approach. The research design logic was to select demonstrably effective teachers as ‘case studies’ in order to understand what effective teachers in the innovation did that promoted greater progress in writing. Qualitative analyses of selected teachers' class sites and students' individual blogs identified features of teaching practice hypothesised to promote student development in writing. To strengthen our understandings, teachers were interviewed to check the comprehensiveness and validity of our interpretation. Classroom observations from these case study teachers were compared with observations from a larger group of teachers to investigate whether identified practices were differentially employed by these effective teachers. Finally, the effects on student writing achievement of the relative presence of these practices in all observed classes were predicted using a hierarchical linear model. Our findings indicate effects of using digital tools in ways that promote complex compositional tasks, discussion and critical thinking. The study adds to a growing number of studies that investigate the nature of effective pedagogy within a digital environment. It contributes to the identification of promising practices for the design of more effective instruction in writing within classes that have ubiquitous digital access.

Synthetic activity: Semiosis, conceptualizations and meaning-making in music composition

Abstract This article is about composition learning of two students in a music programme in upper secondary school with a special regard to how they develop conceptual structures within the musical form of expression and understanding. Semiotics, spectromorphology and cultural historical theory are engaged to discuss the empirical findings that are collected through observation and interview. The compositional task in the study regarded sound-based composition within the field of electroacoustic music, more specifically applying additive synthesis to make raw material for composition, preparatory exercises and finally compositions. Musical meaning-making and compositional creativity were triggered by the encounter of newly learned concepts with internalized and familiar musical thinking. The learning process could be understood in terms of a concept development process (CDP) in language-based learning.

The Influence of Spelling Ability on Vocabulary Choices When Writing for Children With Dyslexia

Spelling is a prerequisite to expressing vocabulary in writing. Research has shown that children with dyslexia are hesitant spellers when composing. This study aimed to determine whether the hesitant spelling of children with dyslexia, evidenced by frequent pausing, affects vocabulary choices when writing. A total of 31 children with dyslexia, mean age 9 years, were compared to typically developing groups of children: the first matched by age, the second by spelling ability. Oral vocabulary was measured and children completed a written and verbal compositional task. Lexical diversity comparisons were made across written and verbal compositions to highlight the constraint of having to select and spell words. A digital writing tablet recorded the writing. Children with dyslexia and the spelling-ability group made a high proportion of spelling errors and within-word pauses, and had a lower lexical diversity within their written compositions compared to their verbal compositions. The age-matched peers demonstrated the opposite pattern. Spelling ability and pausing predicted 53% of the variance in written lexical diversity of children with dyslexia, demonstrating the link between spelling and vocabulary when writing. Oral language skills had no effect. Lexical diversity correlated with written and verbal text quality for all groups. Practical implications are discussed and related to writing models.

Longitudinal Evaluation of the Integration of Digital Musical Instruments into Existing Compositional Work Processes

This paper explores a longitudinal approach to the qualitative evaluation of a set of digital musical instruments, which were developed with a focus on creativity and exploration. The instruments were lent to three electronic musicians/composers for a duration of four weeks. Free exploration periods and a compositional task evaluated using semi-structured show and tell sessions revealed that the context of use had a major influence on how the instruments were experienced by the test persons. Central issues revolved around the learning/adoption process, and around the importance of playability, explorability, and connectivity.

‘A “new music” from nothing’: György Ligeti’s Musica ricercata

The beginning of the 1950s marks a turning-point in György Ligeti’s early career. By that time Ligeti had become disappointed regarding his rather marginal position in Hungarian musical life, and he might well have felt some dissatisfaction with his own artistic output, as well. He recognized that he should leave his former style and build up his own expressive means and musical language from elementary material. For this purpose, he set himself certain compositional tasks, and imposed restrictions on pitch content, intervals, and rhythms ‘as if to build up a “new music” from nothing’. Accordingly, Musica ricercata , which is the first fruit of his experimental project, marks a renewal of Ligeti’s musical thinking primarily on terms of the compositional technique. The present study examines the main problems of compositional technique raised in Musica ricercata (primarily that of chromaticism and dense polyphony) and points out significant influences shown in the work (such as those of Bartók, Stravinsky, and Romanian folklore).

Between Mobility and Stability: Earle Brown's Compositional Process1

Earle Brown's approach to the opposing problems of the exactness and randomness in a musical composition was pioneering and unorthodox. As the article makes clear from comparing Brown's aleatory methods with those of John Cage, Morton Feldman, Pierre Boulez, Witold Lutosławski, and other masters of aleatory music, Brown was in between different directions of aleatory music, and the matter that was behind his ideas about how to organise the musical whole was the problem of balancing between control and flexibility, between mobility and stability. Specifically, Brown's approach to aleatory music was at the staggering intersection between logic and the irrational—in particular, between serial and indeterminate music. Brown never rigorously followed either serialist or aleatory methods in their entirety; however, a mathematical approach and pre-composing structural planning were essential for his compositional process, regardless of the degree of visible freedom and spontaneity his scores provided. The composer carefully and deliberately pre-considered all aspects of the final realisation of the score, and it helped him implement his main compositional tasks that led to multiple performance options. Brown was among the first graduates of the Schillinger House in Boston (later known as the Berklee College of Music). Throughout his life, Brown used Schillinger's musical system as a yardstick to measure his own musical developments and to evaluate the surrounding contemporary musical scene. Using the Schillinger methods of pre-planning and of mathematical calculations for the major music parameters and then allowing the performers to reconfigure the pre-composed materials, Brown was able to create his own ‘sound’, full of shimmering beauty, organic cohesiveness and paradoxical flexibility. The article analyses some of Brown's pre-compositional planning based on Schillinger's principles and also examines Brown's understanding of his primacy in creating open forms.