Atanas Radenski

Enhancing learning management systems to better support computer science education

Many individual instructors -- and, in some cases, entire universities -- are gravitating towards the use of comprehensive learning management systems (LMSs), such as Blackboard and Moodle, for managing courses and enhancing student learning. As useful as LMSs are, they are short on features that meet certain needs specific to computer science education. On the other hand, computer science educators have developed--and continue to develop-computer-based software tools that aid in management, teaching, and/or learning in computer science courses. In this report we provide an overview of current CS specific on-line learning resources and guidance on how one might best go about extending an LMS to include such tools and resources. We refer to an LMS that is extended specifically for computer science education as a Computing Augmented Learning Management System, or CALMS. We also discuss sound pedagogical practices and some practical and technical principles for building a CALMS. However, we do not go into details of creating a plug-in for some specific LMS. Further, the report does not favor one LMS over another as the foundation for a CALMS.

Adapting moodle to better support CS education

Many commercial or open-source systems for organizing courses are available, offering access to course materials, communication support, and receiving and grading student submissions. However, most of these systems are by default not ideally prepared to address specific demands of Computer Science (CS) education. We explore how Moodle as one of the most popular and free systems can be better adapted to support the needs of CS education and provide concrete guidance on features and extensions that could be explored. This report and work based on it can significantly improve courses for educators and students alike

Freedom of choice as motivational factor for active learning

Freedom to choose what, when, and how to contribute in a learning process can motivate students to actively engage and achieve more in their studies. However, freedom of choice complicates course management and may deter instructors from allowing such freedom. Our approach is to utilize existing functionality of course management systems such as Moodle to automatically facilitate and coordinate free student choices and provide much needed relief for instructors at the same time. Using Moodle we have developed novel digital study packs that blend freedom of choice with guidance and control. Our survey shows that assisted freedom of choice is ranked highest in 51% of student responses -- in contrast to unlimited choice at 28% or no choice at all at 21%. Experience reported in this paper may be beneficial for instructors who would like to expand their courses with new motivational learning techniques.

Speeding-up codon analysis on the cloud with local MapReduce aggregation

A notable obstacle to higher performance of data-intensive Hadoop MapReduce (MR) bioinformatics algorithms is the large volume of intermediate data that need to be sorted, shuffled, and transmitted between mapper and reducer tasks. This difficulty manifests itself quite clearly in MR codon analysis which is known to generate voluminous intermediate data that create a bottleneck in basic MR codon analysis algorithms. Our proposed approach to handle the intermediate data bottleneck is local in-mapper aggregation (or simply local aggregation), a technique that helps reduce the intermediate data volume between mapper and reducer tasks in MR. We experimentally evaluate the performance of local aggregation (i) by developing codon analysis MR algorithms with and without local aggregation and (ii) by experimentally measuring their performance on Amazon Web Services (AWS), the Amazon cloud platform. Codon analysis with local aggregation maintains consistently high performance with the growth of larger datasets while basic codon analysis, without local aggregation becomes impractically slow even for smaller datasets. Our results can be beneficial (i) to members of the bioinformatics community who need to perform fast and cost-effective nucleotide MR analysis on the cloud and (ii) to computer scientists who strive to increase the performance of MR algorithms.

Distributed simulated annealing with mapreduce

Simulated annealings high computational intensity has stimulated researchers to experiment with various parallel and distributed simulated annealing algorithms for shared memory, message-passing, and hybrid-parallel platforms. MapReduce is an emerging distributed computing framework for large-scale data processing on clusters of commodity servers; to our knowledge, MapReduce has not been used for simulated annealing yet. In this paper, we investigate the applicability of MapReduce to distributed simulated annealing in general, and to the TSP in particular. We (i) design six algorithmic patterns of distributed simulated annealing with MapReduce, (ii) instantiate the patterns into MR implementations to solve a sample TSP problem, and (iii) evaluate the solution quality and the speedup of the implementations on a cloud computing platform, Amazons Elastic MapReduce. Some of our patterns integrate simulated annealing with genetic algorithms. The paper can be beneficial for those interested in the potential of MapReduce in computationally intensive nature-inspired methods in general and simulated annealing in particular.

Digital support for abductive learning in introductory computing courses

Students who grew up browsing the Web are skilled in what is usually referred to as abduction, a reasoning process that starts with a set of specific observations and then generates the best possible explanation of those observations. In order to exploit the abduction skills of contemporary students, we have developed digital CS1/2 study packs that promote and support active learning through abduction, i.e., abductive learning. The study packs integrate a variety of digital resources: online self-guided labs, e-texts, tutorial links, sample programs, quizzes, and slides. These online packs stimulate students to learn abductively by browsing, searching, and performing self-guided lab experiments. In two years of study pack use, the failure rate in the CS1/2 courses at Chapman University has been reduced from 14% to 5%. The study packs have been published online at studypack.com and adopted in various institutions.

Anomaly-free component adaptation with class overriding

Software components can be implemented and distributed as collections of classes, then adapted to the needs of specific applications by means of subclassing. Unfortunately, subclassing in collections of related classes may require re-implementation of otherwise valid classes just because they utilize outdated parent classes, a phenomenon that is referred to as the subclassing anomaly. The subclassing anomaly is a serious problem since it can void the benefits of component-based programming altogether. We propose a code adaptation language mechanism called class overriding that is intended to overcome the subclassing anomaly. Class overriding does not create new and isolated derived classes as subclassing does, but rather extends and updates existing classes across collections of related classes. If adopted in new languages for component-based programming, or in existing compiled languages such as C# and Java, class overriding can help maintain the integrity of evolving collections of related classes and thus enhance software component adaptability.While other techniques such as reflection and binary code adaptation can be used to reduce the magnitude of the subclassing anomaly, class overriding has the advantage of being easy-to-use and efficient.

A generic all-pairs cluster-computing pipeline and its applications.

In this paper we propose a generic pipeline for all-pairs computations on a cluster of workstations. We use this generic pipeline to derive specific cluster algorithms for three different all-pairs problems: n-body simulation, bubble sort, and Gaussian elimination. We implement the generic pipeline and its derivatives on a cluster of Intel Pentium II workstations using C and the PVM cluster computing environment. We measure and evaluate the performance of the derived algorithms. The n-body and bubble sort algorithms achieve super-linear speedup for large problems.

Using MapReduce Streaming for Distributed Life Simulation on the Cloud

Distributed software simulations are indispensable in the study of large-scale life models but often require the use of technically complex lower-level distributed computing frameworks, such as MPI. We propose to overcome the complexity challenge by applying the emerging MapReduce (MR) model to distributed life simulations and by running such simulations on the cloud. Technically, we design optimized MR streaming algorithms for discrete and continuous versions of Conway’s life according to a general MR streaming pattern. We chose life because it is simple enough as a testbed for MR’s applicability to a-life simulations and general enough to make our results applicable to various lattice-based a-life models. We implement and empirically evaluate our algorithms’ performance on Amazon’s Elastic MR cloud. Our experiments demonstrate that a single MR optimization technique called strip partitioning can reduce the execution time of continuous life simulations by 64%. To the best of our knowledge, we are the first to propose and evaluate MR streaming algorithms for lattice-based simulations. Our algorithms can serve as prototypes in the development of novel MR simulation algorithms for large-scale lattice-based a-life models 1 .

Integrating data-intensive cloud computing with multicores and clusters in an HPC course

This paper presents the design and implementation of a new High-Performance Computing (HPC) course. This course amalgamates the emerging trend of data-intensive cloud computing with the dominant innovation of multicore computing and the important legacy of cluster computing. While others have reported the development of novel HPC courses, we are the first, to our knowledge, to report an upper-level HPC course that introduces the increasingly popular data-intensive cloud computing paradigm - integrally with the better-established shared memory and message-passing models. This course is entirely based on free digital reading resources and open source software. The courses design rationale and implementation - including the use of open source software tools and reading resources - may be beneficial for educators who need to design HPC courses or modules, and especially for those who wish to cover data-intensive cloud computing.