No abstract

Abstract-This paper is an initial approach to create a controller for the game TORCS by learning how another controller or humans play the game. We used data obtained from two controllers and from one human player. The first controller is the winner of the WCCI 2008 Simulated Car Racing Competition, and the second one is a hand coded controller that performs a complete lap in all tracks. First, each kind of controller is imitated separately, then a mix of the data is used to create new controllers. The imitation is performed by means of training a feed forward neural network with the data, using the backpropagation algorithm for learning.

Abstract-Accurate time series forecasting are important for displaying the manner in which the past continues to affect the future and for planning our day to-day activities. In recent years, a large literature has evolved on the use of evolving artificial neural networks (EANNs) in many forecasting applications. Evolving neural networks are particularly appealing because of their ability to model an unspecified non-linear relationship between time series variables. This paper evaluates two methods to evolve neural networks architectures, one carried out with genetic algorithm and a second one carry out with estimation of distribution algorithms. A comparative study between these two methods, with a set of referenced time series will be shown. The object of this study is to try to improve the final forecasting getting an accurate system.

The ability to forecast the future based on past data is a key tool to support individual and organizational decision making. In particular, the goal of Time Series Forecasting (TSF) is to predict the behavior of complex systems by looking only at past patterns of the same phenomenon. In recent years, several works in the literature have adopted Evolutionary Artificial Neural Networks (EANN) for TSF. In this work, we propose a novel EANN approach, where a weighted n-fold validation fitness scheme is used to build an ensemble of neural networks, under four different combination methods: mean, median, softmax and rank-based. Several experiments were held, using six real-world time series with different characteristics and from distinct domains. Overall, the proposed approach achieved competitive results when compared with a non-weighted n-fold EANN ensemble, the simpler 0-fold EANN and also the popular Holt-Winters statistical method.

This paper presents a controller for the 2010 Simulated Car Racing Championship. The idea is not to create the fastest controller but a human-like controller. In order to achieve this, first we have created a process to build a model of the tracks while the car is running and then we used several neural networks which predict the trajectory the car should follow and the target speed. A scripted policy is used for the gear change and to follow the predicted trajectory with the predicted speed. The neural networks are trained with data retrieved from a human player, and are evaluated in a new track. The results shows an acceptable performance of the controller in unknown tracks, more than 20% slower than the human in the same tracks because of the mistakes made when the controller tries to follow the trajectory.