Dimitrios I. Diochnos

On evolvability: the swapping algorithm, product distributions, and covariance

Valiant recently introduced a learning theoretic framework for evolution, and showed that his swapping algorithm evolves monotone conjunctions efficiently over the uniform distribution. We continue the study of the swapping algorithm for monotone conjunctions. A modified presentation is given for the uniform distribution, which leads to a characterization of best approximations, a simplified analysis and improved complexity bounds. It is shown that for product distributions a similar characterization does not hold, and there may be local optima of the fitness function. However, the characterization holds if the correlation fitness function is replaced by covariance. Evolvability results are given for product distributions using the covariance fitness function, assuming either arbitrary tolerances, or a non-degeneracy condition for the distribution and a size bound on the target.

On multiple-instance learning of halfspaces

In multiple-instance learning the learner receives bags, i.e., sets of instances. A bag is labeled positive if it contains a positive example of the target. An @W(dlogr) lower bound is given for the VC-dimension of bags of size r for d-dimensional halfspaces and it is shown that the same lower bound holds for halfspaces over any large point set in general position. This lower bound improves an @W(logr) lower bound of Sabato and Tishby, and it is sharp in order of magnitude. We also show that the hypothesis finding problem is NP-complete and formulate several open problems.

Agent protocols for social computation

Despite the fact that social computation systems involve interaction mechanisms that closely resemble well-known models of agent coordination, current applications in this area make little or no use of the techniques the agent-based systems literature has to offer. In order to bridge this gap, this paper proposes a data-driven method for defining and deploying agent interaction protocols that is entirely based on using the standard architecture of the World Wide Web. This obviates the need of bespoke message passing mechanisms and agent platforms, thereby facilitating the use of agent coordination principles in standard Web-based applications. We describe a prototypical implementation of the architecture and experimental results that prove it can deliver the scalability and robustness required of modern social computation applications while maintaining the expressiveness and versatility of agent interaction protocols.

On the Evolution of Monotone Conjunctions: Drilling for Best Approximations

We study the evolution of monotone conjunctions using local search; the fitness function that guides the search is correlation with Boolean loss. Building on the work of Diochnos and Turan [6], we generalize Valiant’s algorithm [19] for the evolvability of monotone conjunctions from the uniform distribution \({\mathcal U}_n\) to binomial distributions \({\mathcal B}_n\).

SmartSociety -- A Platform for Collaborative People-Machine Computation

Society is moving towards a socio-technical ecosystem in which physical and virtual dimensions of life are intertwined and where people interactions ever more take place with or are mediated by machines. Hybrid Diversity-aware Collective Adaptive Systems (HDA-CAS) is a new generation of socio-technical systems where humans and machines synergetically complement each other and operate collectively to achieve their goals. HDA-CAS introduce the fundamental properties of hybridity and collectiveness, hiding from the users the complexities associated with managing the collaboration and coordination of machine and human computing elements. In this paper we present an HDA-CAS system called Smart Society, supporting computations with hybrid human/machine collectives. We describe the platforms architecture and functionality, validate it on two real-world scenarios involving human and machine elements and present a performance evaluation.

Programming Model Elements for Hybrid Collaborative Adaptive Systems

Hybrid Diversity-aware Collective Adaptive Systems (HDA-CAS) is a new generation of socio-technical systems where both humans and machine peers complement each other and operate collectively to achieve their goals. These systems are characterized by the fundamental properties of hybridity and collectiveness, hiding from users the complexities associated with managing the collaboration and coordination of hybrid human/machine teams. In this paper we present the key programming elements of the Smart Society HDA-CAS platform. We first describe the overall platforms architecture and functionality and then present concrete programming model elements -- Collective-based Tasks (CBTs) and Collectives, describe their properties and show how they meet the hybridity and collectiveness requirements. We also describe the associated Java language constructs, and show how concrete use-cases can be encoded with the introduced constructs.

SmartOrch: an adaptive orchestration system for human-machine collectives

Web-based collaborative systems, where most computation is performed by human collectives, have distinctly different requirements from traditional workflow orchestration systems, as humans have to be mobilised to perform computations and the system has to adapt to their collective behaviour at runtime. In this paper, we present a social orchestration system called SmartOrch, which has been designed specifically for collective adaptive systems in which human participation is at the core of the overall distributed computation. SmartOrch provides a flexible and customisable workflow composition framework that has multi-level optimisation capabilities. These features allow us to manage the uncertainty that collective adaptive systems need to deal with in a principled way. We demonstrate the benefits of SmartOrch with simulation experiments in a ridesharing domain. Our experiments show that SmartOrch is able to respond flexibly to variation in collective human behaviour, and to adapt to observed behaviour at different levels. This is accomplished by learning how to propose and route human-based tasks, how to allocate computational resources when managing these tasks, and how to adapt the overall interaction model of the platform based on past performance. By proposing novel, solid engineering principles for these kinds of systems, SmartOrch addresses shortcomings of previous work that mostly focused on application-specific, non-adaptive solutions.

Leveling-up in heroes of might and magic III

We propose a model for level-ups in Heroes of Might and Magic III, and give an O(1/e2 ln (1/δ)) learning algorithm to estimate the probabilities of secondary skills induced by any policy in the end of the leveling-up process. We develop software and test our model in an experiment. The correlation coefficient between theory and practice is greater than 0.99. The experiment also indicates that the process responsible for the randomization that takes place on level-ups generates only a few different pseudo-random sequences. This might allow exploitation techniques in the near future; hence that process might require reengineering.