## Grapheur: A Software Architecture for Reactive and Interactive Optimization

### BibTeX

@MISC{Brunato_grapheur:a,

author = {Mauro Brunato and Roberto Battiti},

title = {Grapheur: A Software Architecture for Reactive and Interactive Optimization},

year = {}

}

### OpenURL

### Abstract

Abstract This paper proposes a flexible software architecture for interactive multiobjective optimization, with a user interface for visualizing the results and facilitating the solution analysis and decision making process. The architecture is modular, it allows for problem-specific extensions, and it is applicable as a post-processing tool for all optimization schemes with a number of different potential solutions. When the architecture is tightly coupled to a specific problem-solving or optimization method, effective interactive schemes where the final decision maker is in the loop can be developed. An application to Reactive Search Optimization is presented. Visualization and optimization are connected through user interaction: the user is in the loop and the system rapidly reacts to user inputs, like specifying a focus of analysis, or preferences for exploring and intensifying the search in interesting areas. The novelty of the visualization approach consists of using recent online graph drawing techniques, with sampling and mental map preserving schemes, in the framework of stochastic local search optimization. Anecdotal results to demonstrate the effectiveness of the approach are shown for some relevant optimization tasks. 1

### Citations

540 | A heuristic for graph drawing
- Eades
- 1984
(Show Context)
Citation Context ... robust methods can also be considered, for example the BFGS quasi-Newton used in [2]. Kamada-Kawai model [18] — A popular method is that of Kamada and Kawai (KK), inspired by the seminal Eades’ work =-=[8]-=-. The KK energy is: EKK = ∑ (δi,j − di,j) 2 (5) i<j where δi,j is given by the length of the shortest path between two nodes. This definition encourages not only placing at a given distance directly c... |

511 | An Algorithm for Drawing General Undirected Graphs
- Kamada, Kawai
- 1989
(Show Context)
Citation Context .... Gradient descent has well known weaknesses for ill-conditioned problems, and faster and more robust methods can also be considered, for example the BFGS quasi-Newton used in [2]. Kamada-Kawai model =-=[18]-=- — A popular method is that of Kamada and Kawai (KK), inspired by the seminal Eades’ work [8]. The KK energy is: EKK = ∑ (δi,j − di,j) 2 (5) i<j where δi,j is given by the length of the shortest path ... |

468 | Graph drawing by force-directed placement
- Fruchterman, Reingold
- 1991
(Show Context)
Citation Context ...herefore helping to create a more globally consistent display. Only one vertex is moved at each iteration, in a spirit similar to the stochastic gradient descent technique. Fruchterman-Reingold model =-=[10]-=- — Another popular approach has been proposed by Fruchterman and Reingold (FR), who define the FR energy as EFR = 1 3k ∑ i<j d 3 ∑ 2 i,j − k i<j log di,j. (6) After partial derivatives are calculated,... |

252 |
Stochastic Local Search: Foundations & Applications
- Hoos, Stutzle
- 2004
(Show Context)
Citation Context ...to summarize the results. After all, the purpose of mathematical programming is insight, not numbers [11]. The last years have witnessed impressive developments in the area of stochastic local search =-=[14]-=- techniques for combinatorial optimization and in methods at the boundary between optimization and machine learning, a.k.a. intelligent optimization and Reactive Search Optimization (RSO) [4], autonom... |

142 | First-second-order methods for learning: between steepest descent and Newton’s method. Neural Computation
- Battiti
- 1992
(Show Context)
Citation Context ...he above error functions. Gradient descent has well known weaknesses for ill-conditioned problems, and faster and more robust methods can also be considered, for example the BFGS quasi-Newton used in =-=[2]-=-. Kamada-Kawai model [18] — A popular method is that of Kamada and Kawai (KK), inspired by the seminal Eades’ work [8]. The KK energy is: EKK = ∑ (δi,j − di,j) 2 (5) i<j where δi,j is given by the len... |

36 | Online dynamic graph drawing
- Frishman, Tal
- 2008
(Show Context)
Citation Context ... For example, sampling with a bias according to a specific focus and aiming at preserving some relevant graph properties is considered in [22]. The fast drawing of online dynamic graphs is studied in =-=[9]-=-. The issue is of particular relevance when the user is in the loop while an optimization engine is at work: if the graph layout is changing rapidly the focus of attention may be lost unless technique... |

26 |
Effectively Visualizing Large Networks Through
- Rafiei, Curial
- 2005
(Show Context)
Citation Context ...s of scalability required when dealing with very large graphs. For example, sampling with a bias according to a specific focus and aiming at preserving some relevant graph properties is considered in =-=[22]-=-. The fast drawing of online dynamic graphs is studied in [9]. The issue is of particular relevance when the user is in the loop while an optimization engine is at work: if the graph layout is changin... |

15 |
Reactive Search and Intelligent Optimization. Operations research/Computer Science Interfaces
- Battiti, Brunato, et al.
- 2008
(Show Context)
Citation Context ... search [14] techniques for combinatorial optimization and in methods at the boundary between optimization and machine learning, a.k.a. intelligent optimization and Reactive Search Optimization (RSO) =-=[4]-=-, autonomous search [13], instance-aware problem-solving [15]. In many cases, the crucial issue is not that of delivering a single solution, but that of critically analyzing a mass of tentative soluti... |

15 |
The purpose of mathematical programming is insight, not numbers. Interfaces 7(1
- Geoffrion
- 1976
(Show Context)
Citation Context ...a requires appropriate representations. Visualization is crucial to explore optimal solutions, and to summarize the results. After all, the purpose of mathematical programming is insight, not numbers =-=[11]-=-. The last years have witnessed impressive developments in the area of stochastic local search [14] techniques for combinatorial optimization and in methods at the boundary between optimization and ma... |

12 | Parameter adjustment based on performance prediction: Towards an instance-aware problem solver
- Hutter, Hamadi
- 2005
(Show Context)
Citation Context ...n methods at the boundary between optimization and machine learning, a.k.a. intelligent optimization and Reactive Search Optimization (RSO) [4], autonomous search [13], instance-aware problem-solving =-=[15]-=-. In many cases, the crucial issue is not that of delivering a single solution, but that of critically analyzing a mass of tentative solutions, which can easily grow up to thousands or millions. The s... |

11 |
Human-guided simple search: combining information visualization and heuristic search
- Anderson, Anderson, et al.
- 1999
(Show Context)
Citation Context ... mental map are used. Previous work in the area of visualization for optimization includes [21] which discusses the visualization of evolutionary algorithms also through multidimensional scaling, and =-=[1]-=- which deals with human-guided simple search, combining information visualization and heuristic search. In their approach, the computer is responsible only for finding local minima by simple hill-clim... |

10 |
Introduction to Multiobjective Optimization: Interactive Approaches," in Multiobjective Optimization: Interactive and Evolutionary Approaches
- Miettinen, Ruiz, et al.
- 2008
(Show Context)
Citation Context ...nts is called Pareto set (PS). The corresponding set of Pareto optimal objective vectors is called Pareto front (PF). In particular, we focus on interactive multi-objective optimization. According to =-=[20]-=-, through interactive MOO the DM is building a conviction of what is possible and confronting this knowledge with her preferences that also evolve. The assumptions about knowledge and ignorance of the... |

9 | Brain-computer evolutionary multi-objective optimization (BC-EMO): a genetic algorithm adapting to the decision maker
- Battiti, Passerini
- 2010
(Show Context)
Citation Context ...e story and additional opportunities for learning the real preferences exist. A recent proposal of an evolutionary multiobjective optimization algorithm adapting to the decision maker is presented in =-=[5]-=-. In this work, after a brief presentation of the software architecture, we focus on the interactive visualization component. Visualization is here investigated to provide a flexible decision support ... |

5 |
T.: N-to-2-Space Mapping for Visualization of Search Algorithm Performance
- Kadluczka, Nelson, et al.
- 2004
(Show Context)
Citation Context ...ches to quickly and accurately find an optimal solution to a well-defined problem. N-to-2-space mapping for visualization of search algorithm performance by using space-filling curves is discussed in =-=[17]-=-. The main purpose is to discover weaknesses and tune an algorithm. Visualization of Pareto-Sets in Evolutionary MultiObjective Optimization is investigated in [19]: finding a mapping which maintains ... |

4 | Optimal wireless access point placement for location-dependent services
- Battiti, Brunato, et al.
- 2003
(Show Context)
Citation Context ...y might need ad-hoc, domain-dependent definitions. To present anecdotal evidence of the different methods, in the following we present examples related to the Wi-Fi access point placement application =-=[3]-=-. The problem is to install a number of wireless stations (access points, AP) in an area with multiple and conflicting objectives, such as: • the area in which at least one AP signal is above a minimu... |

4 | On effectively finding maximal quasi-cliques in graphs. Learning and Intelligent Optimization
- Brunato, Hoos, et al.
- 2008
(Show Context)
Citation Context ...ansitions between levels that help the user perceive the underlying hierarchical relationship. • k-means clustering, where either the user or the system decide the number of clusters. • Quasi cliques =-=[7]-=- are a recently proposed method for identifying islands of connected components in graphs, where some parameters define the required con-Grapheur: A Software Architecture for Reactive and Interactive... |

3 |
Multicriteria decision making (mcdm): a framework for research and applications
- Bonissone, Subbu, et al.
- 2009
(Show Context)
Citation Context ...es that rely on his intuition, and is able to compare small sets of solutions represented on a two- or at most three-dimensional graph. A recent survey on the topic of multi-objective decision making =-=[6]-=- identifies a number of research and application challenges related to this context: (1) uncertainty in the input and in the evaluation of criteria, (2) high dimensionality of the search and objective... |

2 |
Feature Article–Visualization and Optimization
- Jones
- 1994
(Show Context)
Citation Context ...hemes where the final decision maker (DM for short) is in the loop can be developed. The effectiveness of visualization is well known in the operations research and mathematical programming community =-=[16]-=-. Solving problems involves more than just developing clever algorithms, and giving the optimal solution(s) with minimum f values. Validating and understanding models, algorithms, and data requires ap... |

1 |
optimization, business strategy: a case study
- Visualization
(Show Context)
Citation Context ...ation 5 tifies promising regions of the search space, and intervenes to help it escape nonoptimal local minima. A case study of visualization and optimization for a business strategy is considered in =-=[12]-=-. The visualization capability implicitly allows the user to better formulate the objective function for large optimization runs. As the perceived benefit of different choices does not have an a prior... |

1 |
Visualization of pareto-sets in evolutionary multi-objective optimization
- Koppen, Yoshida
- 2007
(Show Context)
Citation Context ...pace-filling curves is discussed in [17]. The main purpose is to discover weaknesses and tune an algorithm. Visualization of Pareto-Sets in Evolutionary MultiObjective Optimization is investigated in =-=[19]-=-: finding a mapping which maintains most of the dominance relationships is itself a challenging multi-objective optimization task. After a presentation of the overall software architecture, we will co... |

1 |
Visualization of evolutionary algorithms-set of standard techniques and multidimensional visualization
- Pohlheim
- 1999
(Show Context)
Citation Context ...if the graph layout is changing rapidly the focus of attention may be lost unless techniques for preserving a mental map are used. Previous work in the area of visualization for optimization includes =-=[21]-=- which discusses the visualization of evolutionary algorithms also through multidimensional scaling, and [1] which deals with human-guided simple search, combining information visualization and heuris... |