Search

The article discusses two computer science research papers concerning artificial intelligence (AI). Topics explored include the susceptibility of deep convolutional neural networks to input pertubations acknowledged in the 2013 study "Intriguing Properties of Neural Networks," by Christian Szegedy and colleagues, and the capability of sequence-to-sequence language models to execute short computer programs reported in the 2014 study "Learning to Execute," by Wojciech Zaremba and Ilya Sutskever.

The existence of AI application examples is by itself an admirable achievement of engineers and professionals working on them, and certainly there is an interest to move forward to design AI algorithms to make it write proficiently with originality. Artificial Intelligence (AI) has become a trendy topic through advertisements, notes and even warnings about it, claiming that it can be used to answer almost any kind of question. Generating new content requires that AI conduct research work starting from, in principle, human proposed hypothesis, and then getting analyzable results to formulate conclusions. [Extracted from the article]

Reproducibility, the ability to reproduce the results of published papers or studies using their computer code and data, is a cornerstone of reliable scientific methodology. Studies where results cannot be reproduced by the scientific community should be treated with caution. Over the past decade, the importance of reproducible research has been frequently stressed in a wide range of scientific journals such as Nature and Science and international magazines such as The Economist. However, multiple studies have demonstrated that scientific results are often not reproducible across research areas such as psychology and medicine. Statistics, the science concerned with developing and studying methods for collecting, analyzing, interpreting and presenting empirical data, prides itself on its openness when it comes to sharing both computer code and data. In this article, we examine reproducibility in the field of statistics by attempting to reproduce the results in 93 published papers in prominent journals using functional magnetic resonance imaging (fMRI) data during the 2010–2021 period. Overall, from both the computer code and the data perspective, among all the 93 examined papers, we could only reproduce the results in 14 (15.1%) papers, that is, the papers provide both executable computer code (or software) with the real fMRI data, and our results matched the results in the paper. Finally, we conclude with some author-specific and journal-specific recommendations to improve the research reproducibility in statistics. [ABSTRACT FROM AUTHOR]

Over the past two decades, digital flashcards – that is, computer programmes, smartphone apps, and online services that mimic, and potentially improve upon, the capabilities of traditional paper flashcards – have grown in variety and popularity. Many digital flashcard platforms allow learners to make or use flashcards from a variety of sources and customise the way in which flashcards are used. Yet relatively little is known about why and how students actually use digital flashcards during self-regulated learning, and whether such uses are supported by research from the science of learning. To address these questions, we conducted a large survey of undergraduate students (n = 901) at a major U.S. university. The survey revealed insights into the popularity, acquisition, and usage of digital flashcards, beliefs about how digital flashcards are to be used during self-regulated learning, and differences in uses of paper versus digital flashcards, all of which have implications for the optimisation of student learning. Overall, our results suggest that college students commonly use digital flashcards in a manner that only partially reflects evidence-based learning principles, and as such, the pedagogical potential of digital flashcards remains to be fully realised. [ABSTRACT FROM AUTHOR]

Abstract: The paper describes a system for executable papers for publishers enabling them to reuse content and to generate further advances of science and engineering. The executable algorithmic descriptions within a paper are presented in natural language sentences and basic code, thereby making long term compatibility absolute. Authors are required to use publicly numerical libraries on the Internet or references to publications with executable papers. As used by authors the system automatically creates a web of algorithmic knowledge on the Internet. Novelty of new algorithms in publications can be evaluated by automated tools available to authors, reviewers and readers of scientific papers published. [Copyright &y& Elsevier]

Abstract: Adessowiki is a collaborative platform for scientific programming and document writing. It is a wiki environment that carries simultaneously documentation, programming code and results of its execution without any software con- figuration such as compilers, libraries and special tools at the client side. This combination of a collaborative wiki environment, central server and execution of code at rendering time enables the use of Adessowiki as an executable paper platform, since it fulfills the need to disseminate, validate, and archive research data. [Copyright &y& Elsevier]

Having students write programs is the best method to test students programming ability. Most of the time these tests are administered via paper and pencil whereas the homework is completed using a computer and compiler. So why not use a computer and compiler to administer programming tests? This article describes an experiment done in a Java Programming class where both methods of testing were compared. There was no significant difference in students' scores between the two types of testing, but there was a significant difference in the time it took students to complete the test. It took than much longer to complete the test using a computer and compiler. Students were surveyed and they unanimously preferred the computer and compiler method of testing even though it took them longer to complete the test. [ABSTRACT FROM AUTHOR]

In order to assess the extent to which text-mining techniques can be used to gain insights into a particular topic area, we apply hierarchical word clustering and the Term Frequency-Inverse Document Frequency (TF-IDF) measure to articles on computer programming published since the 1970s, when research articles on teaching programming are now more readily available in PDF files. Study 1 compares two sets of papers published before and after the introduction of the concept of Computational Thinking in 2006 to highlight the changes seen in these research sets. Articles mentioned in frequently cited review papers were selected as the target articles to ensure the quality of the sample. Study 2 extends the sample pool to include a range of papers published after the 1970s, allowing us to examine the stability of the conceptual structures identified in Study 1. In both studies, the obtained word clusters or concepts align with known research trends in the programming-education literature. The significance and potential of text-mining techniques are also discussed. [ABSTRACT FROM AUTHOR]

Contribution: This article features a systematic literature review with the objective of presenting a study that reflects the current scenario of research on the cognitive load theory (CLT) in the domain of teaching and learning computer programming. Background: Computer programming is a highly cognitive skill, requiring mastering multiple competencies, and recognized as being difficult to learn, for this reason, the cognitive load (CL) in the learner’s working memory emerged as an influential concept, making CLT one of the most common theories in computing education research. Research Questions: What are the approaches that relate CLT to teaching and learning computer programming? What CLT-related concepts are covered? What evidence is reported with respect to this context? Methodology: Following a formal protocol, a survey was performed for papers linking CLT to teaching and learning programming published between 2010 and 2020. The selection of papers was based on a set of criteria established to drive the selection process, including alignment with the research questions and paper quality assessment. Findings: The approaches applied in the papers are based on measuring the CL; through instructional design based on the development or use of resources or tools, a range of different pedagogy strategies and the CLT concepts. With respect to the concepts, the subjective measurement technique and worked example effect are the most commonly deployed. As far as the evidence reported, the better part is related to the worked example effect and measuring CLs. [ABSTRACT FROM AUTHOR]

A new classification system for computing education papers is presented and applied to every computing education paper published between January 2004 and January 2007 at the two premier computing education conferences in Australia and New Zealand. We find that while simple reports outnumber other types of paper, a healthy proportion of papers address and answer a research question. We find that more papers deal with programming courses than with other courses, and that more than half of all publications are situated in single subjects. To the extent that differing circumstances permit, we compare our results with those of an earlier study of the SIGCSE conference, and find that the Australasian publications include fewer simple reports and more papers describing analysis and experiment. We note a reasonable number of publications on multi-institutional work, which we interpret as evidence of a sense of computing education community within Australia and New Zealand. [ABSTRACT FROM AUTHOR]

This commentary on John Symons' and Jack Horner's paper, besides sharing its main argument, challenges the authors' statement that there is no effective method to evaluate software-intensive systems as a distinguishing feature of software intensive science. It is underlined here how analogous methodological limitations characterise the evaluations of empirical systems in non-software intensive sciences. The authors' claim that formal methods establish the correctness of computational models rather than of the represented programme is here compared with the empirical adequacy problem typifying the model-based reasoning approach in physics, and the remark that testing all the paths of a software-intensive system is unfeasible is related to the enumerative induction problem in the justification of empirical law-like hypotheses in non-software intensive sciences. [ABSTRACT FROM AUTHOR]

Information about several papers discussed at the International Exascale Software Project (IESP) workshop on computer software and hardware related subject is presented. White papers addressing topics, that includes programming models, system software and parallel programming, were submitted by different groups. Information about development in exascale and high speed computing system as discussed in the workshop is also presented.

This paper examines the question of whether computer code could actually substitute for intermediaries, as proponents of distributed ledger technology (DLT) and decentralised finance would argue. Drawing on transaction-cost economics, the paper reviews the role of intermediaries as efficient platforms for the exchange of information, goods and payments. The generic advantage of centralised coordination is compared with the narrative that DLT and decentralised finance could substitute for intermediaries due to their efficiency advantages and the promise of lower costs. The positivist question 'cui bono?' can be applied as a litmus test of these normative narratives in comparison with existing systems. As proposed concepts 'without' intermediaries introduce other intermediaries by the backdoor and no system runs without transaction costs of some sort, the paper concludes that intermediaries cannot be simply replaced. [ABSTRACT FROM AUTHOR]

The recent adoption of deep learning-based models for the processing and coding of multimedia signals has brought noticeable gains in performance, which have established deep learning-based solutions as the uncontested state-of-the-art both for computer vision tasks, targeting machine consumption, as well as, more recently, coding applications, targeting human visualization. Traditionally, applications requiring both coding and computer vision processing require first decoding the bitstream and then applying the computer vision methods to the decompressed multimedia signals. However, the adoption of deep learning-based solutions enables the use of compressed domain computer vision processing, with gains in performance and computational complexity over the decompressed domain approach. For point clouds (PCs), these gains have been demonstrated in the single available compressed domain computer vision processing solution, named Compressed Domain PC Classifier, which processes JPEG Pleno PC coding (PCC) compressed streams using a PC classifier largely compatible with the state-of-the-art spatial domain PointGrid classifier. However, the available Compressed Domain PC Classifier presents strong limitations by imposing a single, specific input size which is associated to specific JPEG Pleno PCC configurations; this limits the compression performance as these configurations are not ideal for all PCs due to their different characteristics, notably density. To overcome these limitations, this paper proposes the first Adaptive Compressed Domain PC Classifier solution which includes a novel adaptive bridge model that allows to process the JPEG Pleno PCC encoded bit streams using different coding configurations, now maximizing the compression efficiency. Experimental results show that the novel Adaptive Compressed Domain PC Classifier allows JPEG PCC to achieve better compression performance by not imposing a single, specific coding configuration for all PCs, regardless of its different characteristics. Moreover, the added adaptability power can achieve slightly better PC classification performance than the previous Compressed Domain PC Classifier and largely better PC classification performance (and lower number of weights) than the PointGrid PC classifier working in the decompressed domain. [ABSTRACT FROM AUTHOR]

Stream reasoning systems are designed for complex decision-making from possibly infinite, dynamic streams of data. Modern approaches to stream reasoning are usually performing their computations using stand-alone solvers, which incrementally update their internal state and return results as the new portions of data streams are pushed. However, the performance of such approaches degrades quickly as the rates of the input data and the complexity of decision problems are growing. This problem was already recognized in the area of stream processing, where systems became distributed in order to allocate vast computing resources provided by clouds. In this paper we propose a distributed approach to stream reasoning that can efficiently split computations among different solvers communicating their results over data streams. Moreover, in order to increase the throughput of the distributed system, we suggest an interval-based semantics for the LARS language, which enables significant reductions of network traffic. Performed evaluations indicate that the distributed stream reasoning significantly outperforms existing stand-alone LARS solvers when the complexity of decision problems and the rate of incoming data are increasing. [ABSTRACT FROM AUTHOR]