Turing’s work on computability served as the founding theoretical framework for computer science. It aimed at formalizing the intuitive notion of algorithm and gave birth to the concepts of logical automaton and abstract machine. The latter are used to evaluate the expressiveness of programming languages. However, current systems are difficult to model within an algorithmic framework and question the relevance of the classical model. The systems we know are indeed characterized by continuous interactions between human agents, physical and computational processes. Programming these interactive systems requires to specify causal relationships between processes, and not simply the execution of computation. In this paper, we propose to define an equivalent of the classical abstract machine, to model programmable interactive systems. Based on our distinction between computation programming and interaction programming and previous theoretical works that have questioned the classical framework, we elicit the minimal requirements of an abstract machine for interaction. This work is a fist step toward an interactive model to evaluate the expressivity of interaction-oriented languages.
17 March 2023 in Journal of Computational and Graphical Statistics
This paper proposes a new Sequential Monte Carlo algorithm to perform online estimation in the context of state space models when either the transition density of the latent state or the conditional likelihood of an observation given a state is intractable. In this setting, obtaining low variance estimators of expectations under the posterior distributions of the unobserved states given the observations is a challenging task. Following recent theoretical results for pseudo-marginal sequential Monte Carlo smoothers, a pseudo-marginal backward importance sampling step is introduced to estimate such expectations. This new step allows to reduce very significantly the computational time of the existing numerical solutions based on an acceptance-rejection procedure for similar performance, and to broaden the class of eligible models for such methods. For instance, in the context of multivariate stochastic differential equations, the proposed algorithm makes use of unbiased estimates of the unknown transition densities under much weaker assumptions than most standard alternatives. The performance of this estimator is assessed for high-dimensional discrete-time latent data models, for recursive maximum likelihood estimation in the context of partially observed diffusion process (POD), and in the case of a bidimensional partially observed stochastic Lotka-Volterra model.
30 December 2022 in Philosophical Problems in Science (Zagadnienia Filozoficzne w Nauce)
Computers are increasingly interactive. They are no more transformational systems producing a final output after a finite execution. Instead, they continuously react in time to external events that modify the course of computing execution. While philosophers have been interested in conceptualizing computers for a long time, they seem to have paid little attention to the specificities of interactive computing. We propose to tackle this issue by surveying the literature in theoretical computer science, where one can find explicit proposals for a model of interactive computing. In that field, the formal modelling of interactive computing systems has been brought down to whether the new interaction models are reducible to Turing Machines. There are three areas where interaction models are framed. The comparison between TMs and interactive system models is at stake in all of them. These areas are namely some works on concurrency by Milner, on Reactive Turing Machines, and on interaction as a new computing paradigm. For each of the three identified models, we present its motivation, sum up its account for interaction and its legacy, and point out issues regarding the understanding of computers. The survey shows difficulties for epistemologists. The reason is that these analyses focus on the formal equivalence between interactive models of computation and classic ones. Such a project is different from addressing how a computing machine can be interactive: in other words, which mechanisms allow it.