In this paper, we present a Monte-Carlo policy rollout technique (called MOCART-CGA) for path planning in dynamic and partially observable real-time environments such as Real-time Strategy games. The emphasis is put on fast action selection motivating the use of Monte-Carlo techniques in MOCART-CGA. Exploration of the space is guided by using corridors which direct simulations in the neighbourhood of the best found moves. MOCART-CGA limits how many times a particular state-action pair is explored to balance exploration of the neighbourhood of the state and exploitation of promising actions. MOCART-CGA is evaluated using four standard pathfinding benchmark maps, and over 1000 instances. The empirical results show that MOCART-CGA outperforms existing techniques, in terms of search time, in dynamic and partially observable environments. Experiments have also been performed in static (and partially observable) environments where MOCART-CGA still requires less time to search than its competitors, but typically finds lower quality plans.
In this paper, coevolution is used to evolve artificial neural networks (ANN) which evaluate board positions of a two player zero-sum game (the virus game). The coevolved neural networks play at a level that beats a group of strong hand-crafted AI players. We investigate the performance of coevolution starting from random initial weights and starting with weights that are tuned by gradient based adaptive learning methods (backpropagation, RPROP and iRPROP). The results of coevolutionary experiments show that pre training of the population is highly effective in this case
This paper investigates the effectiveness of hybrids of learning and evolutionary approaches to find weights and topologies for an artificial neural network (ANN) which is used to evaluate board positions for a two-person zero-sum game, the virus game. Two hybrid approaches: evolutionary RPROP (resilient backpropagation) and evolutionary BP (backpropagation) are described and empirically compared with BP, RPROP, iRPROP (improved RPROP) and evolutionary learning approaches. The results show that evolutionary RPROP and evolutionary BP have significantly better generalisation performance than their constituent learning and evolutionary methods.
The security threats to mobile devices are growing exponentially as the degree of sophistication for these smart devices increase. Mobile devices are not only used for phone calls, they process several enterprise and personal information on daily basis. The comprise to such crucial and private information can cause devastating effects at personal and enterprise levels. In this work, we present a new approach to develop anti-malware programs that can adapt to themselves to the new definitions of the malwares and can detect them in realtime. We propose a machine learning algorithm that can be trained to identify a malicious activity on a smart-phone. We develop a malware and also used existing malwares to generate a training data set to train a malware for detecting the patterns of malicious activities on devices. The paper also presents the key indicators for pattern matching in such cases. The results show that the new definitions of malicious codes are not detectable by existing sophisticated anti-malware programs by using sandboxing techniques. The results also show that all definitions of malwares work under within the same patterns of indicators.