Unsupervised Extraction of Structural Information from High Dimensional Visual Data

We present three unsupervised artificial neural networks for the extraction of structural information from visual data. The ability of each network to represent structured knowledge in a manner easily accessible to human interpretation is illustrated using artificial visual data. These networks are used to collectively demonstrate a variety of unsupervised methods for identifying features in visual data and the structural representation of these features in terms of orientation, temporal and topographical ordering, and stereo disparity.     

Designing a secure e-tender submission protocol

This paper investigates the fundamental difference between a simple e-tender box and a traditional physical tender box, and highlights a series of security traps created by the functional differences. Based on our findings, we have defined the security requirements for an e-tender submission protocol. We also discuss functional limitations of cryptographic technologies. As a result, two secure e-tender submission protocols are proposed which enable a secure e-tender submission. Protocols are assumed to run under the condition that all tendering parties (principal and tenderers) are dishonest players. Our informal and formal security analysis show that these protocols meet their security goals under well known collusion scenarios. Because security is a process not a product, our approach will have broad industry application for developing secure electronic business processes in areas other than e-tendering.

Towards a continuous microfluidic rheometer

In a previous paper we presented a way to measure the rheological properties of complex fluids on a microfluidic chip (Guillot et al., Langmuir 22:6438, 2006). The principle of our method is to use parallel flows between two immiscible fluids as a pressure sensor. In fact, in a such flow, both fluids flow side by side and the size occupied by each fluid stream depends only on both flow rates and on both viscosities. We use this property to measure the viscosity of one fluid knowing the viscosity of the other one, both flow rates and the relative size of both streams in a cross-section. We showed that using a less viscous fluid as a reference fluid allows to define a mean shear rate with a low standard deviation in the other fluid. This method allows us to measure the flow curve of a fluid with less than 250 μL of fluid. In this paper we implement this principle in a fully automated set up which controls the flow rate, analyzes the picture and calculates the mean shear rate and the viscosity of the studied fluid. We present results obtained for Newtonian fluids and complex fluids using this set up and we compare our data with cone and plate rheometer measurements. By adding a mixing stage in the fluidic network we show how this set up can be used to characterize in a continuous way the evolution of the rheological properties as a function of the formulation composition. We illustrate this by measuring the rheological curve of four formulations of polyethylene oxide solution with only 1.3 mL of concentrated polyethylene oxide solution. This method could be very useful in screening processes where the viscosity range and the behavior of the fluid to an applied stress must be evaluated.     

LARS: A learning algorithm for rewriting systems

Whereas there is a number of methods and algorithms to learn regular languages, moving up the Chomsky hierarchy is proving to be a challenging task. Indeed, several theoretical barriers make the class of context-free languages hard to learn. To tackle these barriers, we choose to change the way we represent these languages. Among the formalisms that allow the definition of classes of languages, the one of string-rewriting systems (SRS) has outstanding properties. We introduce a new type of SRS’s, called Delimited SRS (DSRS), that are expressive enough to define, in a uniform way, a noteworthy and non trivial class of languages that contains all the regular languages, , , the parenthesis languages of Dyck, the language of Lukasiewicz, and many others. Moreover, DSRS’s constitute an efficient (often linear) parsing device for strings, and are thus promising candidates in forthcoming applications of grammatical inference. In this paper, we pioneer the problem of their learnability. We propose a novel and sound algorithm (called LARS) which identifies a large subclass of them in polynomial time (but not data). We illustrate the execution of our algorithm through several examples, discuss the position of the class in the Chomsky hierarchy and finally raise some open questions and research directions. This work was supported in part by the IST Program of the European Community, under the PASCAL Network of Excellence, IST-2002-506778. This publication only reflects the authors’ views. Editor: Georgios Paliouras and Yasubumi Sakakibara     

The road to chaos by time-asymmetric Hebbian learning in recurrent neural networks

This letter aims at studying the impact of iterative Hebbian learning algorithms on the recurrent neural network's underlying dynamics. First, an iterative supervised learning algorithm is discussed. An essential improvement of this algorithm consists of indexing the attractor information items by means of external stimuli rather than by using only initial conditions, as Hopfield originally proposed. Modifying the stimuli mainly results in a change of the entire internal dynamics, leading to an enlargement of the set of attractors and potential memory bags. The impact of the learning on the network's dynamics is the following: the more information to be stored as limit cycle attractors of the neural network, the more chaos prevails as the background dynamical regime of the network. In fact, the background chaos spreads widely and adopts a very unstructured shape similar to white noise. Next, we introduce a new form of supervised learning that is more plausible from a biological point of view: the network has to learn to react to an external stimulus by cycling through a sequence that is no longer specified a priori. Based on its spontaneous dynamics, the network decides "on its own" the dynamical patterns to be associated with the stimuli. Compared with classical supervised learning, huge enhancements in storing capacity and computational cost have been observed. Moreover, this new form of supervised learning, by being more "respectful" of the network intrinsic dynamics, maintains much more structure in the obtained chaos. It is still possible to observe the traces of the learned attractors in the chaotic regime. This complex but still very informative regime is referred to as "frustrated chaos."     

Dialogue in context, towards a referential approach in collective learning

In this article, we present research in the making of a collective work environment within the framework of a distance education course. We base our theoretical and methodological standpoints on examples of dialogical discourses recorded within the framework of this CSCL system called Symba. In fact, the results of previous research lead us to rethink our vision of the study of collaborative moments between participants in a computer-supported human learning environment that proposes several communication tools. Redefining the methodological process aiming at finding and understanding these rich learning moments is also necessary. We intend to describe “socio-technical” instances during which these collaboration phases appear. More generally speaking, our aim is to draw up both new theoretical and methodological perspectives that would be reusable in CSCL environments; in view of the nature of these two perspectives, and the diversity of the domain knowledge (sociology, cognitivism, linguistics, philosophy, statistics, etc.) brought to bear in the study of the environment in question, our approach constitutes a trans-disciplinary reassessment of the uses of the communication tools—and the study thereof—proposed.

Fair,biased, and self-balancing merge operators: Their specification and implementation in Concurrent Prolog

The problem of allowing a dynamically changing set of processes fair access to a shared resource is considered, in the context of communication-stream based systems. It is argued that fair binary merge operators alone cannot solve this problem satisfactorily. Two solutions are proposed. One employs binary merge operators with a programmable bias; the other binary and ternary fair merge operators capable of self-balancing, using the concept of 2–3 trees. A Concurrent Prolog implementation of these operators is described. The implementation of the self-balancing merge operators illustrates the expressive power of incomplete messages, a programming technique that supports messages that contain communication channels as arguments. In the course of implementing the self-balancing merge operator, it was necessary to develop a distributed variant of the 2–3 tree deletion algorithm.     

Wafer-level fabrication process for fully encapsulated micro-supercapacitors with high specific energy

In this paper a wafer-level process is proposed to fully integrate carbon-based micro-supercapacitor onto silicon substrate. This process relies on the deposition of a paste containing carbon, PVDF and acetone into cavities etched in silicon. After electrolyte deposition in a controlled atmosphere, a wafer-level encapsulation is realized. Cyclic voltammetry performed on non-encapsulated micro-components showed specific energy of 257 mJ cm⁻² for 336 μm deep cavities. The specific encapsulation process developed was tested separately and proved to be efficient in terms of resistance to organic electrolytes and mechanical strength.