Francesco Maiorana

Mining massive datasets by an unsupervised parallel clustering on a GRID: Novel algorithms and case study

This paper proposes three novel parallel clustering algorithms based on the Kohonens SOM aiming at preserving the topology of the original dataset for a meaningful visualization of the results and for discovering associations between features of the dataset by topological operations over the clusters. In all these algorithms the data to be clustered are subdivided among the nodes of a GRID. In the first two algorithms each node executes an on-line SOM, whereas in the third algorithm the nodes execute a quasi-batch SOM called MANTRA. The algorithms differ on how the weights computed by the slave nodes are recombined by a master to launch the next epoch of the SOM in the nodes. A proof outline demonstrates the convergence of the proposed parallel SOMs and provides indications on how to select the learning rate to outperform both the sequential SOM and the parallel SOMs available in the literature. A case study dealing with bioinformatics is presented to illustrate that by our parallel SOM we may obtain meaningful clusters in massive data mining applications at a fraction of the time needed by the sequential SOM, and that the obtained classification supports a fruitful knowledge extraction from massive datasets.

Discovering Genes-Diseases Associations From Specialized Literature Using the Grid

This paper proposes a novel method for text mining on the Grid, aimed at pointing out hidden relationships for hypothesis generation and suitable for semi-interactive querying. The method is based on unsupervised clustering and the outputs are visualized with contextual information. Grid implementation is crucial for feasibility. We demonstrate it with a mining run for discovering genes-diseases associations from bibliographic sources and annotated databases. The proposed methodology is in view of a Grid architecture specialized in bioinformatics mining tasks. Some performance considerations are provided.

An Evaluation of Cellular Neural Networks for the Automatic Identification of Cephalometric Landmarks on Digital Images

Several efforts have been made to completely automate cephalometric analysis by automatic landmark search. However, accuracy obtained was worse than manual identification in every study. The analogue-to-digital conversion of X-ray has been claimed to be the main problem. Therefore the aim of this investigation was to evaluate the accuracy of the Cellular Neural Networks approach for automatic location of cephalometric landmarks on softcopy of direct digital cephalometric X-rays. Forty-one, direct-digital lateral cephalometric radiographs were obtained by a Siemens Orthophos DS Ceph and were used in this study and 10 landmarks (N, A Point, Ba, Po, Pt, B Point, Pg, PM, UIE, LIE) were the object of automatic landmark identification. The mean errors and standard deviations from the best estimate of cephalometric points were calculated for each landmark. Differences in the mean errors of automatic and manual landmarking were compared with a 1-way analysis of variance. The analyses indicated that the differences were very small, and they were found at most within 0.59 mm. Furthermore, only few of these differences were statistically significant, but differences were so small to be in most instances clinically meaningless. Therefore the use of X-ray files with respect to scanned X-ray improved landmark accuracy of automatic detection. Investigations on softcopy of digital cephalometric X-rays, to search more landmarks in order to enable a complete automatic cephalometric analysis, are strongly encouraged.

Automatic landmarking of cephalograms by cellular neural networks

Cephalometric analysis is a time consuming measurement process by which experienced orthodontist identify on lateral craniofacial X-rays landmarks that are needed for diagnosis and treatment planning and evaluation. High speed and accuracy in detection of craniofacial landmarks are widely demanded. A prototyped system, which is based on CNNs (Cellular Neural Networks) is proposed as an efficient technique for landmarks detection. The first stage of system evaluation assessed the image output of the CNN, to verify that it included and properly highlighted the sought landmark. The second stage evaluated performance of the developed algorithms for 8 landmarks. Compared with the other methods proposed in the literature, the findings are particularly remarkable with respect to the accuracy obtained. Another advantage of a CNN based system is that the method can either be implemented via software, or directly embedded in the hardware, for real-time performance.

Feeding back learning resources repurposing patterns into the “information loop”: opportunities and challenges

The paper outlines a model for framing the representation and treatment of information gathered from the reuse and repurposing of learning resources from distributed repositories. The model takes into account as sources of information both static user-edited or automatically generated metadata fields and the emerging, dynamic information clouds that surrounds a learning resource when users comment on it, tags it, or explicitly links it to other learning resources. By coordinating these separate information layers, the advantages that can be achieved are reducing the semantic gap occurring when unanticipated contexts of use are to be described by resorting only to predefined vocabularies; and improvements in the relevance of the retrieved resources after a query. To achieve this “coordination” it is proposed that the textual descriptions of the repurposing activity with respect to the intended learning outcomes and pedagogical strategies are fed to a dynamic unsupervised classification method that operates on the above mentioned information spaces, and that supports exploratory search by suggesting associations. It is argued that the proposed analogical retrieval, as opposed to standard query matching, is more fit to tracking the loci of innovation and sustaining the formation of best practices in the community.

Cellular neural networks and dynamic enhancement for cephalometric landmarks detection

Cephalometric landmarks detection is a knowledge intensive activity to identify on X-rays of the skull key points to perform measurements needed for medical diagnosis and treatment. We have elsewhere proposed CNNs (Cellular Neural Networks) to achieve an accuracy in automated landmarks detection suitable for clinical practice, and have applied the method for 8 landmarks located on the bone profile. This paper proposes and evaluates a CNNs approach augmented by local image dynamic enhancemet for other 3 landmarks that are notoriously difficult to locate; the advantages of this method in the landmark detection problem are pointed out.

New Horizons in the Assessment of Computer Science at School and Beyond: Leveraging on the ViVA Platform

A revolution is taking place across Europe and worldwide in how we teach our children about computing, in primary and secondary school. Out goes ICT and how to use Microsoft Office; in comes coding and computer science. Assessment has a crucial role to play in this revolution. If teachers use low-quality assessment instruments we will end-up teaching the wrong subject; and viceversa. This paper reviews the state of the field, and makes concrete, achievable proposals for developing shared, high quality assessments for computer science. Central to this proposal is the collaborative platform VIVA (the Vilnius collaboratively coded and Validated computer science questions/tasks for Assess- ment). Two requirements are key to VIVA: 1) support for multiple competency frameworks, so that the contributors can meta-tag resources with respect to the framework they are most familiar with; and 2) support for crowdsourcing the validation of each question/task and its mapping to competencies. The use of a taxonomy of questions/tasks type that has been mapped to computational thinking concepts and to a competency framework is proposed. Some seed questions are already available in the online platform prototype, and various supporters have granted permission to use large questions banks. The design requirements of a full implementation of the VIVA platform for a modern and effective approach to assessment including support for digital badges, are outlined; and some preliminary results from a survey administered to the initial contributors to VIVA are presented.

Input Noise Robustness and Sensitivity Analysis to Improve Large Datasets Clustering by Using the GRID

In this paper we investigate the performance of a refined version of the Kohonen self organizing feature maps algorithm in terms of classification correctness when we inject in a sparse input matrix different kinds of noise and compared these classification results with the one without noise. The analysis not only gives indications on the classification errors due to noisy data, but also let a methodology to emerge in order to identify the portion of the input matrix that must be controlled with great care for avoiding classification errors. The methodology also suggests a suitable data partitioning approach for a GRID implementation of the described algorithm. The methodological indications were successfully verified by a case study belonging to the bioinformatics field.

Teaching algorithms: Visual language vs flowchart vs textual language

There is a strong movement asserting the importance of quality education all over the world and for students of all ages. Many educators believe that in order to achieve this 21st century skills must be taught and that digital literacy should be coupled with rigorous Computer Science principles and computational thinking. Accordingly this work will describe a didactic experience in an introductory programming course by describing the context, pedagogical approach, content of the course based on a procedure-first approach, technologies used, research questions addressed, experimental design adopted, data collection and analysis and the main conclusion supported by qualitative and quantitative data. The research questions focus on understanding which is the best medium to design algorithms by comparing flow chart and the Scratch programming language and by evaluating whether using textual language is worth the effort of the syntactic burden imposed by these languages. An analysis of quantitative and qualitative data revealed that both a visual programming and a flow-chart approach are suitable for algorithm design with no statistical difference in terms of number of errors and time taken to write the corresponding code in a textual language. However, the high number of errors suggest that using visual programming allows the student to focus on the problem solving activities.

Quizly: A live coding assessment platform for App Inventor

There is a strong worldwide movement which is pushing for the teaching of serious computer science principles besides reading, writing and basic numeracy starting from first grade and reaching all students across all grades. This is being done through both formal initiatives carried out by international organizations and at the national level by putting forward curricula, some of which are mandatory. In order to accomplish this goal, and based on the consensus that computer science is not programming and that programming languages are a tool, visual languages have become the preferred method for teaching introductory courses in computer science. The absence of rigid syntactic rules makes them the ideal tool for focusing on problem solving and computational thinking activities. Recent reports have pointed out the need for supporting the international community of teachers by providing assessment methods, an internationally validated question repository as well as tools and assessment platforms. In this context our work presents an assessment platform for formative, summative and informal assessment of computer science competencies by using a visual language, namely App Inventor, which allows for the rapid development of a mobile app and has a strong appeal to the younger generation of students. The capability to log user activity allows the teacher to monitor the progression in the students learning path as well as her/his solution-building approach.