Sofea

  • Published on
    28-Aug-2014

  • View
    242

  • Download
    12

Embed Size (px)

DESCRIPTION

 

Transcript

<ul><li> POOL VS. ISLAND BASED EVOLUTIONARY ALGORITHMS: AN INITIAL EXPLORATIONJ.J. Merelo1, A.M. Mora1, C.M. Fernandes1, Anna Esparcia-Alczar2,J.L.J. Laredo3University of Granada1S2 Grupo2University of Luxembourg3 </li> <li> Scope EAs are optimization algorithms Typically running in single sequential computers OR Parallel symetric computers Proof of Concept (PoC) Pool-based EAs (Shared Memory: here CouchDB) OneMax: Evaluation time (small) Concurrent access (big) Small number of clients </li> <li> Outline Introduction Evolutionary Algorithms (EAs) Pool and Island EAs on CouchDB (SofEA) CouchDB Analysis of results Improving baseline SofEA: Island vs. Elite Scalability Conclusions </li> <li> IntroductionBio-inspired optimization methods based on population 1 0 0 1 1 1 1 1 1 1 f (x) 25 1 0 1 1 0 0 1 0 0 1 f (x) 14 EA1 0 0 1 1 1 0 1 1 1 f (x) 10 1 0 0 0 1 0 0 0 0 0 f (x) 8 SofEA 1 0 0 0 0 0 0 1 1 1 f (x) 4 CouchDB </li> <li> IntroductionBio-inspired optimization methods based on population Selection 1 0 0 1 1 1 1 1 1 1 f (x) 25 1 0 1 1 0 0 1 0 0 1 f (x) 14 EA1 0 0 1 1 1 0 1 1 1 f (x) 10 1 0 0 0 1 0 0 0 0 0 f (x) 8 SofEA 1 0 0 0 0 0 0 1 1 1 f (x) 4 CouchDB </li> <li> IntroductionBio-inspired optimization methods based on population Selection Recombination 1 0 0 1 1 1 1 1 1 1 f ( x ) 25 1 0 1 1 0 0 1 0 0 1 f ( x ) 14 EA1 0 1 1 0 1 1 1 1 1 f ( x ) 30 SofEA 1 0 1 1 0 1 0 1 1 1 f ( x ) 29 CouchDB </li> <li> Introduction Persistance BaseSofEA Resilience Asynchrony CouchDB1 Step EA Block EA SofEA CouchDB </li> <li> Introduction Persistance BaseSofEA Resilience Asynchrony CouchDB1 Step EA Block EliteSofEA CouchDB Reducing the design space 1 Step EA EA Block SofEA CouchDB </li> <li> Introduction Persistance BaseSofEA Resilience Decreasing Communications Asynchrony CouchDB IslandSofEA1 Step EA Block CouchDB EliteSofEA 25 Steps EA Block CouchDB Reducing the design space 1 Step EA EA Block SofEA CouchDB </li> <li> Introduction Document-oriented STORE Stores Key-Documents pairs Map/Reduce requests using JavaScript API RESTfull with JSON Olainberg at Flickr EA SofEA CouchDB </li> <li> Analysis of results Running times: EliteSofEA vs. BaseSofEA Block size: BaseSofEA: 128 EliteSofEA: 128, 32 Improving baseline SofEA: Island vs. Elite Scalability </li> <li> Analysis of results 1 to 8 clients Population 256 Algorithmic performance: Evaluation-to-solutions: EliteSofEA vs. IslandSofEA Improving baseline SofEA: Island vs. Elite Scalability </li> <li> Analysis of results Running times: EliteSofEA vs. IslandSofEA Speed-up: EliteSofEA vs. IslandSofEA EliteSofEA accesses 25 more times CouchDB Algorithm 1 Client 8 Clients Speed-up (1x256) (8x32) EliteSofEA 30.43 12.19 2.5 IslandSofEA 2.067 1.173 1.7 Improving baseline SofEA: Island vs. Elite Scalability </li> <li> Conclusions SofEA: Persistent, Asynchronous, Resilient pool-based EA PoC: It finds the optimum, it scales, it works!! Future works: Limit of scalability on time-consuming problems Latency/computation decreases </li> <li> THANKS FOR YOUR ATTENTION!! </li> </ul>