|
| |
Model Selection
This project is dedicated to stimulate research and reveal the state-of-the art in "model selection" by organizing a competition followed by a workshop. Model selection is a problem in statistics, machine learning, and data mining. Given training data consisting of input-output pairs, a model is built to predict the output from the input, usually by fitting adjustable parameters. Many predictive models have been proposed to perform such tasks, including linear models, neural networks, trees, and kernel methods. Finding methods to optimally select models, which will perform best on new test data, is the object of this project. The competition will help identifying accurate methods of model assessment, which may include variants of the well-known cross-validation methods and novel techniques based on learning theoretic performance bounds. Such methods are of great practical importance in pilot studies, for which it is essential to know precisely how well desired specifications are met.
Predictive Uncertainty in Environmental Modeling
This competition aims to compare methods for evaluating estimates of predictive uncertainty in regression problems arising in environmental modelling. Often, in addition to providing accurate predictions, we must also provide an indication of the uncertainty of our predictions (which arise both due to the noise inherent in the data and due to estimating the parameters of a model from a finite sample of data). This may take the form of error bars, giving the mean and variance of a Gaussian predictive distribution or a set of quantiles describing the likely values for the target. The competition is based on real-world environmental modelling tasks, which are quite noisy, and where the noise processes involved may be non-Gaussian. The competition is based on the recent Pascal predictive uncertainty challenge and is intended to complement the special session on Computational Intelligence in Earth and Environmental Sciences at WCCI-2006. The challenge website is now open.
|
|
| |
Othello
The goal of this competition is to supply the winning evaluation (utility) function for Othello (also known as Reversi). The scientific aims of the competition are to further our understanding of which type of architecture is best for performing Othello position evaluation, of which learning methods work best, and the level of play that can be achieved at 1-ply.
Binary Series Prediction
The prediction of time series is a standard application of evolutionary computation. This contest asks the competitors to predict a binary time series. The contestant is given 10,000 bits from a time series as training data and asked to predict the next 10,000 bits of time series data one at a time.
Evolved Art
Evolved art is art produced by any evolutionary process. We expect most entries to result from some form of evolutionary computation. Visualizations of evolutionary computation are also potential sources of evolutionary art. As long as your entry can be defended as being generated by an evolutionary process we will accept it, space and other rules permitting.
Nonlinear Programming
The competitor must submit a program which will be tested on a number of unknown non-linear programming problems. The winner will be determined based on the consistency in locating the optimum and the time required to do so.
Huygens Probe Competition
In this competition you will be given access to a series of 20 "moons"- fractal landscapes (generated by sequences of meteor impacts) that are wrapped in both x and y dimensions. For each moon you will be allowed 1000 probes (evalutations) to find the lowest point on the surface that you can. You may use any technique at your disposal.
|
|
Back to top |
|
 |
