Dimitrios Milios

Smoothed model checking for uncertain Continuous-Time Markov Chains

We consider the problem of computing the satisfaction probability of a formula for stochastic models with parametric uncertainty. We show that this satisfaction probability is a smooth function of the model parameters under mild conditions. This enables us to devise a novel Bayesian statistical algorithm which performs model checking simultaneously for all values of the model parameters from observations of truth values of the formula over individual runs of the model at isolated parameter values. This is achieved by exploiting the smoothness of the satisfaction function: by modelling explicitly correlations through a prior distribution over a space of smooth functions (a Gaussian Process), we can condition on observations at individual parameter values to construct an analytical approximation of the function itself. Extensive experiments on non-trivial case studies show that the approach is accurate and considerably faster than naive parameter exploration with standard statistical model checking methods.

U-Check: Model Checking and Parameter Synthesis Under Uncertainty

Novel applications of formal modelling such as systems biology have highlighted the need to extend formal analysis techniques to domains with pervasive parametric uncertainty. Consequently, machine learning methods for parameter synthesis and uncertainty quantification are playing an increasingly significant role in quantitative formal modelling. In this paper, we introduce a toolbox for parameter synthesis and model checking in uncertain systems based on Gaussian Process emulation and optimisation. The toolbox implements in a user friendly way the techniques described in a series of recent papers at QEST and other primary venues, and it interfaces easily with widely used modelling languages such as PRISM and Bio-PEPA. We describe in detail the architecture and use of the software, demonstrating its application on a case study.

Studying Emergent Behaviours in Morphogenesis Using Signal Spatio-Temporal Logic

Pattern formation is an important spatio-temporal emergent behaviour in biology. Mathematical models of pattern formation in the stochastic setting are extremely challenging to execute and analyse. Here we propose a formal analysis of the emergent behaviour of stochastic reaction diffusion systems in terms of Signal Spatio-Temporal Logic, a recently proposed logic for reasoning on spatio-temporal systems. We present a formal analysis of the spatio-temporal dynamics of the Bicoid morphogen in Drosophila melanogaster, one of the most important proteins in the formation of the horizontal segmentation in the development of the fly embryo. We use a recently proposed framework for statistical model checking of stochastic systems with uncertainty on parameters to characterise the parametric dependence and robustness of the French Flag pattern, highlighting non-trivial correlations between the parameter values and the emergence of the patterning.

Probabilistic Programming Process Algebra

Formal modelling languages such as process algebras are widespread and effective tools in computational modelling. However, handling data and uncertainty in a statistically meaningful way is an open problem in formal modelling, severely hampering the usefulness of these elegant tools in many real world applications. Here we introduce ProPPA, a process algebra which incorporates uncertainty in the model description, allowing the use of Machine Learning techniques to incorporate observational information in the modelling. We define the semantics of the language by introducing a quantitative generalisation of Constraint Markov Chains. We present results from a prototype implementation of the language, demonstrating its usefulness in performing inference in a non-trivial example.

Efficient Stochastic Simulation of Systems with Multiple Time Scales via Statistical Abstraction

Stiffness in chemical reaction systems is a frequently encountered computational problem, arising when different reactions in the system take place at different time-scales. Computational savings can be obtained under time-scale separation. Assuming that the system can be partitioned into slow- and fast- equilibrating subsystems, it is then possible to efficiently simulate the slow subsystem only, provided that the corresponding kinetic laws have been modified so that they reflect their dependency on the fast system. We show that the rate expectation with respect to the fast subsystem’s steady-state is a continuous function of the state of the slow system. We exploit this result to construct an analytic representation of the modified rate functions via statistical modelling, which can be used to simulate the slow system in isolation. The computational savings of our approach are demonstrated in a number of non-trivial examples of stiff systems.

Global Optimization of Analogy-Based Software Cost Estimation with Genetic Algorithms

Estimation by Analogy is a popular method in the field of software cost estimation. A number of research approaches focus on optimizing the parameters of the method. This paper proposes an optimal global setup for determining empirically the best method parameter configuration based on genetic algorithms. We describe how such search can be performed, and in particular how spaces whose dimensions are of different type can be explored. We report results on two datasets and compare with approaches that explore partially the search space. Results provide evidence that our method produces similar or better accuracy figures with respect to other approaches.

Policy Learning for Time-Bounded Reachability in Continuous-Time Markov Decision Processes via Doubly-Stochastic Gradient Ascent

Continuous-time Markov decision processes are an important class of models in a wide range of applications, ranging from cyber-physical systems to synthetic biology. A central problem is how to devise a policy to control the system in order to maximise the probability of satisfying a set of temporal logic specifications. Here we present a novel approach based on statistical model checking and an unbiased estimation of a functional gradient in the space of possible policies. The statistical approach has several advantages over conventional approaches based on uniformisation, as it can also be applied when the model is replaced by a black box, and does not suffer from state-space explosion. The use of a stochastic gradient to guide our search considerably improves the efficiency of learning policies. We demonstrate the method on a proof-of-principle non-linear population model, showing strong performance in a non-trivial task.

A genetic algorithm approach to global optimization of software cost estimation by analogy

Estimation by Analogy is a popular method in the field of software cost estimation. However, the configuration of the method affects estimation accuracy, which has a great effect on project management decisions. This paper proposes an optimal global setup for determining empirically the best parameter configuration based on genetic algorithms. Those parameters involve the definition of project similarity, the number of analogies and the way of adjusting the analogies used. We describe how such a search can be performed in the parameter space spanned by these parameters, which are essentially of different type. We report results on two datasets and compare with approaches that explore partially the search space. Results provide evidence that our method produces similar or better accuracy figures with respect to other approaches.

Compositional approximate markov chain aggregation for PEPA models

Approximate Markov chain aggregation involves the construction of a smaller Markov chain that approximates the behaviour of a given chain. We discuss two different approaches to obtain a nearly optimal partition of the state-space, based on different notions of approximate state equivalence. Both approximate aggregation methods require an explicit representation of the transition matrix, a fact that renders them inefficient for large models. The main objective of this work is to investigate the possibility of compositionally applying such an approximate aggregation technique. We make use of the Kronecker representation of PEPA models, in order to aggregate the state-space of components rather than of the entire model.

Probabilistic model checking for continuous time Markov chains via sequential Bayesian inference

Probabilistic model checking for systems with large or unbounded state space is a challenging computational problem in formal modelling and its applications. Numerical algorithms require an explicit representation of the state space, while statistical approaches require a large number of samples to estimate the desired properties with high confidence. Here, we show how model checking of time-bounded path properties can be recast exactly as a Bayesian inference problem. In this novel formulation the problem can be efficiently approximated using techniques from machine learning. Our approach is inspired by a recent result in statistical physics which derived closed-form differential equations for the first-passage time distribution of stochastic processes. We show on a number of non-trivial case studies that our method achieves both high accuracy and significant computational gains compared to statistical model checking.