Search

Qu’est-ce qui fait que les logiques machiniques résistent à l’utilisateur ? Il semble que l’on prête aux machines un potentiel utopique d’unification de toutes les formes de logiques vers un seul et unique modèle binaire. Pourtant, la transparence, postulée par cette utopie unificatrice, s’oppose à la situation technique effective des machines, révélant la tension existante entre l’opacité de la matérialité du hardware et l’idéal utopique du software. En réduisant l’information au strict minimum, les machines se doteraient d’un pouvoir computationnel infini. Héritière de la machine universelle théorique de Turing, cette machine utopique serait capable de tout et de rien. Si tout est traitable par la computation, alors la machine devient véritablement universelle en rendant possible un langage unifié dans des circuits de logiques électroniques. Le design des programmes se situe à l’intermédiaire entre la logique machinique et la production de programmes fonctionnels. Ce texte explore les modalités techniques de la traduction de la logique au sein des machines dans le cadre du design de programmes. Nous émettons l’hypothèse que la stratification des couches successives de logiciels, entre le designer et l’électronique, impliquent de multiples traductions. Cet empilement permet à la machine d’exécuter le programme tout en rendant la compréhension des logiques machiniques difficile pour l’utilisateur, une traduction bruitée vers un langage inintelligible. Les logiques de substitution d’une chaîne de caractères par une autre chaîne de caractères partagent des traits communs avec l’analyse du cryptage de l’information, avec la cryptanalyse. Si les modèles analytiques sont proches, les objectifs linguistiques diffèrent : du point de vue de la machine, la traduction, automatisée, s’attache à la syntaxe, et du point de vue de l’humain, elle s’attache particulièrement à la sémantique. Cette distinction fondamentale contribue au sentiment d’opacification des logiques et à la résistance des machines face à l’entendement. Or, ces procédés imposent une non-réversibilité de la traduction : interdépendance entre les strates, compilation, obfuscation, enjeux économiques, constituent autant d’instances qui semblent ériger la dissimulation des logiques en dogme technique, au profit d’une idéologie fonctionnelle restreignant paradoxalement l’intelligibilité des machines computationnelles. What makes machine logic so difficult to understand for users ? It seems as if we attribute to machines the utopian potential of unifying all forms of logic into one single binary model. Yet, the transparency proposed by this utopian unification contradicts the actual technical situation of machines, revealing an existing tension between the material opacity of hardware and the utopian ideal of software. By reducing information to its bare minimum, machines would acquire infinite computational power. Heir to Turing’s theoretical universal machine, this utopian machine would be capable of everything and nothing. If computation can process everything, then the machine becomes truly universal by enabling a unified language embedded in electronic logic circuits. Program design is at the interface between machine logic and the production of functional programs. This essay explores the technical modalities of logic translations inherent to machines within the context of program design. We hypothesize that the successive layers of software, between a designer and the electronics, implies successive translations. This layering allows the machine to execute the program while simultaneously making it difficult for the end user to understand the machine-based logic, which would be equivalent to a noisy translation resulting in an unintelligible language. The logic of substituting one string (sequence of characters) for another string shares common traits with information encryption analysis, ergo cryptanalysis. While the analytical models are similar, the linguistic goals are quite different : from the machine standpoint, automated translation focuses on syntax, and from a human standpoint, it focuses particularly on semantics. This fundamental distinction contributes to a perception of machine logic opacity and makes it all the more difficult for the machine’s functioning to be understood. These processes reinforce the non-reversibility of translation : interdependence between layers, compilation, obfuscation, economic stakes, are all instances that seem to provide the basis of concealment of logic as technical dogma, favoring a functional ideology that paradoxically restricts the intelligibility of computational machines.

We study a natural variant of the implicational fragment of propositional logic. Its formulas are pairs of conjunctions of positive literals, related together by an implicational-like connective; the semantics of this sort of implication is defined in terms of a threshold on a conditional probability of the consequent, given the antecedent: we are dealing with what the data analysis community calls confidence of partial implications or association rules. Existing studies of redundancy among these partial implications have characterized so far only entailment from one premise and entailment from two premises, both in the stand-alone case and in the case of presence of additional classical implications (this is what we call "relative entailment"). By exploiting a previously noted alternative view of the entailment in terms of linear programming duality, we characterize exactly the cases of entailment from arbitrary numbers of premises, again both in the stand-alone case and in the case of presence of additional classical implications. As a result, we obtain decision algorithms of better complexity; additionally, for each potential case of entailment, we identify a critical confidence threshold and show that it is, actually, intrinsic to each set of premises and antecedent of the conclusion.

M.Sc. (Mathematics) This dissertation describes the solution toa specific logical problem, the satisfiability problem, in a logic of games called Alternating-time Temporal Logic (ATL). Computation Tree Logic (CTL) is a discrete branching-time temporal logic for reasoning about labelled transition systems. ATL extends CTL to describe gametheoretic situations, where multiple agents together determine the evolution of the system. In particular, ATL explicitly provides for describing the abilities of coalitions of agents in such systems. Weprovide an automata-based decision procedure for ATL by translating the satisfiability problem for an ATL formula to the nonemptiness problem for an Alternating Biichi 'free Automaton. The key result that enables this translation is a oundedbranching tree model theorem for ATL, proving that a satisfiable formula is also satisfiable in a tree model of bounded branching degree. In terms of complexity, we show that satisfiability in ATL is complete for exponential time, which agrees with the corresponding complexity result for the fragment CTL. Closely related to ATL is an independently developed family of modal logics, the Coalition Logics. The presented results also provide a satisfiability procedure for Extended Coalition Logic interpreted over strongly playable coalition models. The structure of the dissertation is as follows: • Chapter 1 is an introduction to the topic, provides an overview of the results and a preview of the dissertation. • Chapter 2 presents some mathematical preliminaries regarding trees, automata, fixed points and game theory. • Chapter 3 discusses CTL and in particular an automata-based satisfiability procedure for CTL. • Chapter 4 introduces Alternating-time Temporal Logic (ATL) as a logic of games. • Chapter 5 contains the main results of the dissertation: first we prove a boundedbranching tree model property for ATL. Then the construction of the required automaton for satisfiability checking is described. • Chapter 6 relates the present work to some other logics of games, and in particular the Coalition Logics. • Chapter 7 finalises the dissertation with a conclusion and a look at some future research directions that might be pursued following the present work.

This thesis presents an approach to image classification via a Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) on the SRC-6 reconfigurable computer for use in classifying Low Probability of Intercept (LPI) radar emitters. The rationale behind the previously unexplored use of new reconfigurable computers combined with neural networks for this application is the potential for near real-time classification. Current potential near-peer competitors have access to LPI technology, so development of quick classification methods is crucial for ships to determine intent and to enable the possibility for self-defense against these types of emitters. The neural network, based on work conducted by Professor Phillip E. Pace of the Naval Postgraduate School (NPS), generates integer-cast weights by first using a sequential processor to conduct floating-point backpropagation to train the network on potential timefrequency images that allows generation of weights with lower overall Root Mean Squared (RMS) errors. The weights are then used in a parallel-processing reconfigurable computer for close to real-time classification. A second method of direct pixel comparison using Exclusive-Or (XOR) logic is presented as an alternative image classification method. Comparisons to similar representations in C++ are provided, for use in judging comparative error levels and timing between parallel and sequential processing methods. http://archive.org/details/neuralnetworkdes109452396 Approved for public release; distribution is unlimited.

We present a second-order logic of proportional quantifiers, SOLP, which is essentially a first-order language extended with quantifiers that act upon second-order variables of a given arity r and count the fraction of elements in a subset of r-tuples of a model that satisfy a formula. Our logic is capable of expressing proportional versions of different problems of complexity up to NP-hard as, for example, the problem of deciding if at least a fraction 1/n of the set of vertices of a graph form a clique; and fragments within our logic capture complexity classes as NL and P, with auxiliary ordering relation. When restricted to monadic second-order variables, our logic of proportional quantifiers admits a semantic approximation based on almost linear orders, which is not as weak as other known logics with counting quantifiers (restricted to almost orders), for it does not have the bounded number of degrees property. Moreover, we show that, in this almost-ordered setting, different fragments of this logic vary in their expressive power, and show the existence of an infinite hierarchy inside our monadic language. We extend our inexpressibility result of almost-ordered structure to a fragment of SOLP, which in the presence of full order captures P. To obtain all our inexpressibility results, we developed combinatorial games appropriate for these logics, whose application could go beyond the almost-ordered models and hence are interesting by themselves.