Sign boundary and hand articulation feature recognition in Sign Language videos

Machine Translation - In this paper we present a recommendation system for (semi-)automatic annotation of sign language videos exploiting deep learning techniques, which handle handshape...     

Using distributed ledger technology to democratize neural network training

Artificial Intelligence has regained research interest, primarily because of big data. Internet expansion, social networks and online sensors led to the generation of an enormous amount of information daily. This unprecedented data availability boosted Machine Learning. A research area that has greatly benefited from this fact is Deep Neural Networks. Nowadays many use cases require huge models with millions of parameters and big data are proven to be essential to their proper training. The scientific community has proposed several methods to generate more accurate models. Usually, these methods need high performance infrastructure, which limits their applicability to large organizations and institutions that have the required funds. Another source of concern is privacy; anyone using the leased processing power of a remote data center, must trust another entity with their data. Unfortunately, in many cases sensitive data were leaked, either for financial exploitation or due to security issues. However, there is a lack of research studies when it comes to open communities of individuals with commodity hardware, who wish to join forces in a way that is non-binding and without the need for a central authority. Our work on LEARNAE attempts to fill this gap, by creating a way of providing training in Artificial Neural Networks, featuring decentralization, data ownership and fault tolerance. This article adds some important pieces to the puzzle: It studies the resilience of LEARNAE when dealing with network disruptions and proposes a novel way of embedding low-energy sensors that reside in the Internet of Things domain, retaining at the same time the established distributed philosophy.

     

MOLES: A New Approach to Modeling the Environmental and Economic Impacts of Urban Policies

Computational Economics - This paper presents the Multi-Objective Local Environmental Simulator (MOLES), an urban Computable General Equilibrium model with selected microsimulation features that...     

Occlusion Culling Algorithms: A Comprehensive Survey

In this paper, occlusion culling techniques that appeared in the last decade are reviewed. Occlusion culling techniques are responsible for reducing the polygons rendered by the graphics hardware with the target of achieving real-time rendering. The various techniques are discussed in detail and a synopsis table with their main characteristics is given.     

Case Studies and Information Technology in Civil Engineering Learning

A multimedia-supported case study is presented that deals with a large-scale civil engineering project. This case study is based on the complete analysis, design, and construction files of the developer of the project. The multimedia instructional tool has integrated different disciplines and input from the project manager, senior students, and instructors of different disciplines. This multimedia tool includes hypertext links to modular and stratified information. Information is made available through the user's control in a logical, interactive, deductive, and disciplinary way that follows the flow of a civil engineering project. Text, graphics, and videos are included. The multimedia tool presents a teaching tool that enables senior class civil engineering students to learn about planning, design, and construction phases of a civil engineering project, as well as their integration. The in-class use of the tool and project discussions help the students to relate better their previous academic knowledge to “real-life” problems, enhance their creativity, and increase the level of retention of the new knowledge acquired. Future enhancements and learning possibilities are discussed with the use of additional capabilities of information technology. The multimedia application developed has been used in an academic environment; however, with some modifications, it could be used to improve the “learning curve” of new employees in a company environment.     

Energetically and pressure driven phase transitions in molecularly thin films of n-alkanes

Molecular dynamics simulations were carried out on molecularly thin films of n-octane confined between topographically smooth solid surfaces. We focused on determining the effect of increasing solid surface-methylene unit energetic affinity and the effect of increasing pressure (normal load) of the film in inducing liquid-solid phase transitions. Simulations of films wide enough to accomodate three segmental layers showed an abrupt transition in the structural features at a critical value of the characteristic energy that quantified the affinity between solid surfaces and methylene units. This energetically driven transition was evident from the discontinuous increase of intermolecular order, a precipitous extension of the octane molecules and freezing of molecular migration and rotation. Increasing pressure had a similar effect in inducing a liquid-solid phase transition. The characteristics of the transition showed that it is a mild first-order transition from a highly ordered liquid to a poorly organized solid. These findings demonstrate that the solidification of nanoscopically thin films of linear alkanes is a general phenomenon (driven either energetically or by increasing pressure), and does not require the aid of commensurate surface topography. Our findings on relatively wider films (5 segmental diameters wide) show that the interfacial layer undergoes a similar first-order phase transition with increasing solid-methylene unit energetic affinity. This energy threshold is significantly higher than the one observed in thin film simulations.     

Deep-Space Transport Protocol: A novel transport scheme for Space DTNs

The Delay-/Disruption-Tolerant Networking Architecture calls for new design principles that will govern data transmission and retransmission scheduling over challenged environments. In that context, novel routing, transport and application layer algorithms have to be established in order to achieve efficient and reliable communication between DTN-nodes.In this study, we focus on the evolution of the terrestrial Internet into the Interplanetary or Space Internet and propose adoption of the Deep-Space Transport Protocol (DS-TP) as the transport layer scheme of choice for the space networking protocol stack. We present DS-TP’s basic design principles and we evaluate its performance both theoretically and experimentally. We verify that practice conforms with theory and observe great performance boost, in terms of file delivery time between DTN-nodes, in case of DS-TP. In particular, the gain of DS-TP against conventional proposals for deep-space communications increases with the link error rate; under conditions DS-TP can improve the performance of the transport layer protocol by a factor of two (i.e., DS-TP can become two times faster than conventional protocols).     

On the performance of two-hop message spreading in DTNs

In this paper, a Delay Tolerant Network environment is considered where the source is in full control of the two-hop spreading mechanism by setting key parameters such as the number of copies allowed to be spread in the network and the delay bound of the messages. The introduced analysis allows for a differentiation between the source of the message and the intermediate nodes (in terms of e.g. transmission power, speed or cooperation degree). Analytical expressions for the cumulative distribution function (cdf) of the delivery delay and the induced overhead are extracted, taking into account the fact that the source node may continue spreading copies after the message delivery. In addition, a fairly accurate approximate expression for the cdf of the delivery delay is also derived and validated through simulations.     

A simplified nonlinear stability analysis of an imperfect rectangular two-bar frame

A simplified nonlinear stability analysis for moderately large rotations and small strains is performed on a rectangular two-bar frame subjected to a concentrated eccentrically applied joint load. Such a simplification consisting of adopting linear kinematic relations leads to very reliable results for the initial postbuckling path in the vicinity of the critical point of the above imperfect frame. The existence of an asymmetric bifurcation point is thoroughly discussed and a direct evaluation of the bifurcational load is readily obtained. Using this technique the effect of imperfection sensitivity is also addressed. A qualitative analysis associated with the physical phenomenon yields a substantial reduction of the computational work. The efficiency and reliability of this approximate nonlinear stability analysis proposed herein is illustrated by means of several examples for which a lot of numerical results based on a more accurate nonlinear analysis are available.