Invited Speakers


Kenneth Jong
Kenneth A. De Jong
Title: Evolutionary Computation: A Unified Approach (Tutorial) - Sun 21/10 14:00
Abstract: The field of Evolutionary Computation has experienced tremendous growth over the past 15 years, resulting in a wide variety of evolutionary algorithms and applications.  The result poses an interesting dilemma for many practitioners in the sense that, with such a wide variety of algorithms and approaches, it is often hard to  se the relationships between them, assess strengths and weaknesses, and make good choices for new application areas.
    This tutorial is intended to give an overview of a general EC framework that can help compare and contrast approaches, encourages crossbreeding, and facilitates intelligent design choices.   The use of this framework is then illustrated by showing how traditional EAs can be compared and contrasted with it, and how new EAs can be effectively designed using it.
    Finally, the framework is used to identify some important open issues that need further research.
Title: Evolutionary Computation: Where We are and Where We're Headed (Talk) - Wed 24/10 8:30
    Abstract: The field of evolutionary computation has experienced a significant growth of interest and activity in the past decade.  This has resulted in fresh perspectives and a flurry of new results in both theory and applications.  This presentation will summarize the recent progress and characterize some of the remaining unresolved research issues.




Thomas
                  Dietterich
Thomas G. Dietterich
Title: Machine Learning in Ecological Science and Environmental Policy (Tutorial) - Sun 21/10 16:30
Abstract: There are many exciting opportunities for machine learning research in ecological science and environmental policy. This tutorial will present examples of existing and emerging applications of machine learning and challenges for machine learning research.  The tutorial is organized around a "pipeline" from data collection through the development of machine learning models to the optimization of management policies.  The tutorial is divided into three main sections.

Section 1: Data acquisition.  We describe the many kinds of sensors (including human observers) that are employed in ecological research and describe three case studies of the application of machine learning: optimal sensor placement, computer vision for species monitoring, and machine learning for analysis of Doppler weather radar.

Section 2: Machine Learning of Ecological Models. We first describe static models of species distribution with three case studies: Species Distribution Models (SDMs) from presence-only data, SDMs eBird (bird watcher) data, and SDMs that also model detection and observer expertise.  Second, we describe dynamical models including Collective Graphical Models for species migration and Meta-Population Models.

Section 3: Policy Optimization. This section describes four case studies and associated machine learning research problems: linking populations of Red Cockaded Woodpeckers (an endangered species), robust optimal solution of Markov Decision Processes (MDPs) for fisheries management, solution of spatio-temporal MDPs for wildfire management, and application of Partially-Observable Markov Decision Processes for managing difficult-to-observe invasive species. 
Title: Machine Learning and Computational Sustainability (Talk) - Mon 22/10 8:30
Abstract: Computational Sustainability is the study of computer science methods for solving sustainability challenges.  These challenges include sustainable cities, sustainable development, energy, transportation, ecological science, and ecosystem management.  In this talk, I will focus on research opportunities in ecological science and ecosystem management and describe some of the work in my lab on these problems. This includes (a) automated data cleaning and anomaly detection in sensor data streams, (b) species distribution modeling including modeling of bird migration from citizen science data, and (c) design of optimal policies for managing wildfires and invasive species.  The novel machine learning challenges include flexible anomaly detection for multiple data streams, explicit models of sampling bias and measurement processes, combining probabilistic graphical models with non-parametric learning methods, collective graphical models for reasoning with count data, and optimization of complex spatio-temporal Markov decision processes.


Anthony Hunter
Anthony Hunter
Title: Introduction to Computational Models of Argument (Tutorial) - Sun 21/10 14:00 (slides[pdf])
Abstract: Computational Models of Argument are being developed with the aim of reflecting how human argumentation uses conflicting information to construct and analyse arguments. Argumentation involves identifying arguments and counterarguments relevant to an issue (e.g. What are the pros and cons for the safety of mobile phones for children?). Argumentation may also involve weighing, comparing, or evaluating arguments (e.g. What sense can we make of the arguments concerning mobile phones for children?) and it may involve drawing conclusions (e.g. A parent answering the question "Are mobile phones safe for my children?"). In addition, argumentation may involve convincing an audience (e.g. A politician making the case that mobile phones should be banned for children because the risk of radiation damage is too great). In this tutorial, we will consider both graph-based and logic-based formalizations of argumentation, introducing some of the basic concepts, and reviewing a range of proposals and results.
Title: Computational Models of Dialogical Argumentation (Talk) - Tue 23/10 8:30 (slides[pdf])
    Abstract: Computational models of dialogical argumentation aim to reflect how humans use and resolve conflicting information, opinions, beliefs, etc. by exchanging and analysing arguments. This may be as part of a wider process such as decision making, sense making, or negotiation. In these formal models, a dialogue between two or more agents is normally made up of a set of communicative acts, called moves, a set of rules that state which moves are legal to make at any point in a dialogue (the protocol), and a set of rules that define the effect of making a move. In addition, each agent may have a private strategy that specifies which of the allowed moves it should do in the current state of the dialogue given its goals. In this talk, we will consider some of the main approaches to modelling dialogical argumentation, and identify some of the many research questions that have been identified.

Scott Sanner
Scott Sanner
Title: Recent Advances in Continuous Planning (Tutorial) - Sun 21/10 16:30
Abstract: This tutorial will outline the importance of continuous variables and actions in deterministic and stochastic planning and how one can cope with the complexity of planning for such problems in practice. The first part of the tutorial will focus on modeling real-world phenomena both abstractly (deterministic transition models, MDPs, POMDPs) and in terms of formal planning languages (PDDL, PPDDL, and RDDL).
The second part of the tutorial will focus on solution methods, ranging from search in deterministic settings, through to approximation methods such as discretization and sampling.From this point, the tutorial will shift to an in-depth focus on very recent work that allows the exact solution of an expressive range of continuous planning problems via a technique referred to as Symbolic Dynamic Programming. The tutorial will conclude by briefly connecting continuous planning to related fields such as Control Theory and Scheduling.
Title: Data Structures for Efficient Inference and Optimization in Expressive Continuous Domains (Talk) - Tue 23/10 16:30 (slides[pdf])
    Abstract: To date, our ability to perform exact closed-form inference or optimization with continuous variables is largely limited to special well-behaved cases.  This talk argues that with an appropriate representation and data structure, we can vastly expand the class of models for which we can perform exact, closed-form inference. This enables novel solutions to many problems of interest in AI, machine learning, operations research and beyond, some of which have been unsolved for 50+ years.
This talk is in two parts.  In the first part, I introduce an extension of the algebraic decision diagram (ADD) to continuous variables -- termed the extended ADD (XADD) -- to represent arbitrary piecewise functions (nb, arbitrary pieces, not just hyper-rectangular) over discrete and continuous variables and show how to define and efficiently compute elementary arithmetic operations, integrals, and maximization for various restrictions of these functions.  In the second part, I briefly cover a wide range of novel applications where the XADD may be applied: (a) exact inference in expressive discrete and continuous variable graphical models, (b) factored, parameterized linear and quadratic optimization (a generalization of LP and QP solving), (c) exact solutions to piecewise convex functions that enable a number of novel applications in machine learning, and (d) exact solutions to continuous state, action, and observation sequential decision-making problems -- which includes closed-form solutions to previously unsolved problems in operations research. This is joint work with Zahra Zamani & Ehsan Abbasnejad (Australian National University), Karina Valdivia Delgado & Leliane Nunes de Barros (University of Sao Paulo), and Simon Fang (M.I.T.).


Alan Kirman
Alan Kirman
Title: Individual and Collective Rationality in Complex Systems (Talk) - Mon 22/10 16:30
Abstract: This talk will analyse the emergence of collective behaviour which has a certain rationality from the point of view of the aggregate but which does not correspond to that of the individual. In economics it has long been recognised that, as a result of “externalities” behaviour that is rational from the point of view of the individual may lead to collective disasters as in the case of the “Tragedy of the Commons”. However, even more interesting are the cases where individuals who interact locally and have only limited information can together produce outcomes which have regularities which could not have been predicted from the individual motives. These may be beneficial but may also, as a result of contagion, lead the system to sudden collapse. Modelling this sort of phenomena usually involves starting with a formal analysis of a very simple case and then simulating a more complex and realistic system to see whether the results for the simple case still hold in the more general case. I will illustrate this with a number of examples, from some observations about the analogies with social insects, and ant and bees in particular, to recent work on the extension of the Schelling model of segregation. In the latter we include preferences for neighbourhood income as well as racial composition and introduce a housing market. We analyse segregation in terms of both race and income and the relation between house prices and these factors. I will also discuss a model of the market for mortgage backed securities and show how the rational reactions of a few individuals to a small change in the probability of default on the underlying assets can lead to a collapse in the price of the assets. In each case the self organisation of a complex adaptive system produces aggregate results which could not have been predicted from the analysis of a “representative agent”.


Leandro de Castro
Leandro de Castro
Title: Natural Computing: The Grand Challenges and Two Case Studies (Talk) - Wed 24/10 16:30
Abstract: Although the origins of Natural Computing (NC) can be traced back to the early days of Computer Science in the 1940s, it only became a formal discipline around the year 2000. Since then, there has been a flood of research papers on the use of Nature to inspire the design of novel algorithms. Now, it is felt that the field faces some Grand Challenges so as to progress and mature. A new perspective of the field has emerged, broadening its scope, and three Grand Challenges have been proposed: to transform Natural Computing into a transdisciplinary science; to harness information processing in natural systems; and to formally and consistently engineer Natural Computing approaches. This talk introduces and discusses the Grand Challenges in Natural Computing Research and then, for showing the pragmatism and usefulness of NC, follows with the presentation of two real-world applications of Natural Computing, one for Social Media and another in e-Commerce.








Sponsored by

Organized by

Supported by