Agent-based simulations, adaptive algorithms and solvers (ABS-AAS) Session 3

Time and Date: 16:40 - 18:20 on 6th June 2016

Room: Macaw

Chair: Maciej Paszynski

60 Efficient Memetic Continuous Optimization in Agent-based Computing [abstract]
Abstract: This paper deals with a concept of memetic search in agent-based evolutionary computation. In the presented approach, local search is applied during mutation of an agent. Using memetic algorithms causes increased demand on the computing power as the number of fitness function calls increases, therefore careful planning of the fitness computing (through the proposed local search mechanism based on caching parts of the fitness function) leads to significant lowering of this demand. Moreover, applying local search with care, can lead to gradual improvement of the whole population. In the paper the results obtained for selected high-dimensional (5000 dimensions) benchmark functions are presented. Results obtained by the evolutionary and memetic multi-agent systems are compared with classic evolutionary algorithm.
Wojciech Korczynski, Aleksander Byrski, Marek Kisiel-Dorohinicki
107 Reinforcement Learning with Multiple Shared Rewards [abstract]
Abstract: A major concern in multi-agent coordination is how to select algorithms that can lead agents to learn together to achieve certain goals. Much of the research on multi-agent learning relates to reinforcement learning (RL) techniques. One element of RL is the interaction model, which describes how agents should interact with each other and with the environment. Discrete, continuous and objective-oriented interaction models can improve convergence among agents. This paper proposes an approach based on the integration of multi-agent coordination models designed for reward-sharing policies. By taking the best features from each model, better agent coordination is achieved. Our experimental results show that this approach improves convergence among agents even in large state-spaces and yields better results than classical RL approaches.
Douglas M. Guisi, Richardson Ribeiro, Marcelo Teixeira, André Pinz Borges, Fabrício Enembreck
353 Why invasive Argentine ant supercolonies are a limited social transition [abstract]
Abstract: Around the world, invasive Argentine ants have formed “supercolonies”: societies of societies where separate colonies share one identity and behave as one colony, but can they be maintained long enough to transform into the next major evolutionary transition? Here we argue no, and use an agent-based, spatially-explicit model of Argentine ant supercolonies to demonstrate how the extreme supercolony structure of invasive Argentine ants will collapse over time. Our model explains supercolonies as ethnocentric collections of ant colonies, making social decisions based on cuticular hydrocarbon (CHC) markers demarcating in-group from out-group. When simulated, we observe supercolonies as fragile assemblages, where the shared identity that constitutes supercolonies depends on consistent warfare, and fades over time in invasive habitats where supercolonies face no competitors. With unicoloniality in ants being a rare and phylogenetically scattered trait, they appear to be only a temporary evolutionary phenomenon. Our model explains why. It is because the consistency of “self” cannot be maintained at the scale of supercolonies, and therefore the next major evolutionary transition must be found elsewhere.
Brian Whyte, John Marken, Matthew Zeffermen, Nicole Rooks and Keenan Mack
294 Hybrid direct and iterative solver with library of multi-criteria optimal orderings for h adaptive finite element method computations [abstract]
Abstract: In this paper we present a multi-criteria optimization of element partition trees and resulting orderings for multi-frontal solver algorithms executed for two dimensional h adaptive finite element method. In particular, the problem of optimal ordering of elimination of rows in the sparse matrices resulting from adaptive finite element method computations is reduced to the problem of finding of optimal element partition trees. Given a two dimensional h refined mesh, we find all optimal element partition trees by using the dynamic programming approach. An element partition tree defines a prescribed order of elimination of degrees of freedom over the mesh. We utilize three different metrics to estimate the quality of the element partition tree. As the first criterion we consider the number of floating point operations (FLOPs) performed by the multi -frontal solver. As the second criterion we consider the number of memory transfers (MEMOPS) performed by the multifrontal solver algorithm. As the third criterion we consider memory usage ( NONZEROS) of the multi-frontal direct solver. We show the optimization results for FLOPs vs MEMOPS as well as for the execution time estimated as FLOPs+100*MEMOPS vs NONZEROS. We obtain Pareto fronts with multiple optimal trees, for each mesh, and for each refinement level. We generate a library of optimal elimination trees for small grids with local singularities. We also propose an algorithm that for a given large mesh with multiple local singularities finds separators that partition the mesh into multiple sub -grids, each one with local singularity. We compute Schur complements over the local singularities using the optimal trees from the library, and later we submit the sequence of Schur complements into the iterative solver ILUPCG.
Hassan Aboueisha, Konrad Jopek, Bartłomiej Medygrał, Szymon Nosek, Mikhail Moshkov, Anna Paszynska, Maciej Paszynski, Keshav Pingali