Patricia Anthony

PRICAI 2002: Trends in Artificial Intelligence

Systems biology is an attempt to understand biological system as system thereby triggering innovations in medical practice, drug discovery, bio-engineering, and global sustainability problems. The fundamental difficulties lies in the complexity of biological systems that have evolved through billions of years. Nevertheless, there are fundamental principles governing biological systems as complex evolvable systems that has been optimized for certain environmental constraints. Broad range of AI technologies can be applied for systems biology such as text-mining, qualitative physics, marker-passing algorithms, statistical inference, machine learning, etc. In fact, systems biology is one of the best field that AI technologies can be best applied to make high impact research that can impact real-world. This talk addresses basic issues in systems biology, especially in systems drug discovery and coral reef systems biology, and discusses how AI can contribute to make difference. P. Anthony, M. Ishizuka, and D. Lukose (Eds.): PRICAI 2012, LNAI 7458, p. 1, 2012. c

Malay named entity recognition based on rule-based approach

A Named-Entity Recognition (NER) is part of the process in Text Mining and it is a very useful process for information extraction. This NER tool can be used to assist user in identifying and detecting entities such as person, location or organization. However, different languages may have different morphologies and thus require different NER processes. For instance, an English NER process cannot be applied in processing Malay articles due to the different morphology used in different languages. This paper proposes a Rule-Based Named-Entity Recognition algorithm for Malay articles. The proposed Malay NER is designed based on a Malay part-of-speech (POS) tagging features and contextual features that had been implemented to handle Malay articles. Based on the POS results, proper names will be identified or detected as the possible candidates for annotation. Besides that, there are some symbols and conjunctions that will also be considered in the process of identifying named-entity for Malay articles. Several manually constructed dictionaries will be used to handle three named-entities; Person, Location and Organizations. The experimental results show a reasonable output of 89.47% for the F-Measure value. The proposed Malay NER algorithm can be further improved by having more complete dictionaries and refined rules to be used in order to identify the correct Malay entities system.

A comparative investigation of non-linear activation functions in neural controllers for search-based game AI engineering

The creation of intelligent video game controllers has recently become one of the greatest challenges in game artificial intelligence research, and it is arguably one of the fastest-growing areas in game design and development. The learning process, a very important feature of intelligent methods, is the result of an intelligent game controller to determine and control the game objects behaviors’ or actions autonomously. Our approach is to use a more efficient learning model in the form of artificial neural networks for training the controllers. We propose a Hill-Climbing Neural Network (HillClimbNet) that controls the movement of the Ms. Pac-man agent to travel around the maze, gobble all of the pills and escape from the ghosts in the maze. HillClimbNet combines the hill-climbing strategy with a simple, feed-forward artificial neural network architecture. The aim of this study is to analyze the performance of various activation functions for the purpose of generating neural-based controllers to play a video game. Each non-linear activation function is applied identically for all the nodes in the network, namely log-sigmoid, logarithmic, hyperbolic tangent-sigmoid and Gaussian. In general, the results shows an optimum configuration is achieved by using log-sigmoid, while Gaussian is the worst activation function.

Pareto evolution and co-evolution in cognitive game AI synthesis

The Pareto-based Differential Evolution (PDE) algorithm is one of the current state-of-the-art Multi-objective Evolutionary Algorithms (MOEAs). This paper describes a series of experiments using PDE for evolving artificial neural networks (ANNs) that act as game-playing agents. Three systems are compared: (i) a canonical PDE system, (ii) a co-evolving PDE system (PCDE) with 3 different setups, and (iii) a co-evolving PDE system that uses an archive (PCDE-A) with 3 different setups. The aim of this study is to provide insights on the effects of including co-evolutionary techniques on a well-known MOEA by investigating and comparing these 3 different approaches in evolving intelligent agents as both first and second players in a deterministic zero-sum board game. The results indicate that the canonical PDE system outperformed both co-evolutionary PDE systems as it was able to evolve ANN agents with higher quality game-playing performance as both first and second game players. Hence, this study shows that a canonical MOEA without co-evolution is desirable for the synthesis of cognitive game AI agents.

PageRank: A modified random surfer model

PageRank is an approach to evaluate the importance of a web page implemented by Google. It is one of the important system features that Google usedin order to improve the quality of search result in the original version of Google apart from utilizing link (anchor) in web pages. The success of Google has led to various researches on the theory behind Google search. PageRank is one of the theories that arestudied over the years by researchers. Various theories are proposed to enhance PageRank in terms of its quality andcomputation time. This paper explains the behavior of Markov chain involved in a random surfer model from the original PageRank. A modified random surfer model is proposed, which could lead to a more predictable time for computing PageRank.

Reserve price strategy for seller agent in multiple simultaneous auctions

Internet auction is popular due to the flexibility and convenience that it offers to consumers. In online auctions, sellers are confronted with the dilemma of deciding the best reserve price for the items to be auctioned off. In an auction site such as eBay, one can always find the same item being sold by multiple sellers in different auctions. Determining this reserve price is not a straightforward decision process due to the complexity and vagueness of the online auction environment. Setting the reserve price too high may result in no sale whilst setting the reserve price too low may result in a sale with low profit. The main focus of this paper is to analyze the performance of the selling agents with varying pricing strategy when offering item for auctions. The strategy could be categorized as low, intermediate, high, and random based on its sellers types. Our study showed that the use of these strategies produced significant drawbacks with negative impacts towards the overall selling upshots. To counteract these shortcomings, we develop an autonomous seller agent using a heuristic decision making framework. To derive the best reserve price, several constraints are considered including the number of competitors, the number of bidders, the auction length, and the profit that the seller desires. This paper presents the design, implementation and evaluation of our selling algorithm that a seller agent can use for auctioning homogeneous goods among multiple overlapping English auctions. In this work, we modeled our market simulation for a single unit auction using an independent private value framework with dynamic participation entry.

Meta search engine powered by DBpedia

The evolution of information retrieval technology on the web has led to the idea of semantic search engine in which it understands the meaning and the context of the search query. As a consequence, the search results returned by this type of search engine should match closely with the query. However, the Web is still dominated by Web 2.0 in which information and data is presented in an unstructured manner and is only fit for human consumption. Hence, building a semantic search engine is a very challenging task and there is still a lot of improvement that needs to be done to achieve the desirable results. As an example, if we search for “food that is not halal”, existing semantic search engines still ignore the term of “not” resulting in inaccurate search. In view of this problem, this paper proposes a semantic meta search engine that utilizes the power of a traditional search engine (Google) and enriches the search result using DBpedia as the knowledge base to produce better results. This paper also describes the application of the knowledge base contained in DBpedia to deliver an improved search engine.

Agent for Predicting Online Auction Closing Price in a Simulated Auction Environment

Auction markets provide centralized procedures for the exposure of purchase and sale orders to all market participants simultaneously. Online auctions have effectively created a large marketplace for participants to bid and sell products and services over the Internet. eBay pioneered the online auction in 1995. As the number of demand for online auction increases, the process of monitoring multiple auction houses, picking which auction to participate in, and making the right bid become a challenging task for the consumers. Hence, knowing the closing price of a given auction would be an advantage since this information will be useful and can be used to ensure a win in a given auction. However, predicting a closing price for an auction is not easy since it is dependent on many factors. This paper reports on a predictor agent that utilises the Grey System Theory to predict the closing price for a given auction. The performance of this predictor agent is compared with another well known technique which is the Artificial Neural Network. The effectiveness of these models is evaluated in a simulated auction environment.

Neural network ensembles for video game AI using evolutionary multi-objective optimization

Recently, there has been an increasing interest in game artificial intelligence (AI). Game AI is a system that makes the game characters behave like human beings that is able to make smart decisions to achieve the target in a computer or video game. Thus, this study focuses on an automated method of generating artificial neural network (ANN) controller that is able to display good playing behaviors for a commercial video game. In this study, we create neural-based game controller for screen-capture of Ms. Pac-Man using a multi-objective evolutionary algorithm (MOEA) for training or evolving the architectures and connection weights (including biases) in ANN corresponding to conflicting goals of minimizing complexity in ANN and maximizing Ms. Pac-man game score. In particular, we have chosen the commonly-used Pareto Archived Evolution Strategy (PAES) algorithm for this purpose. After the entire training process is completed, the controller is tested for generalization using the optimized networks in single network (single-net) and neural network ensemble (multi-net) environments. The multi-net model is compared to single-net model, and the results reveal that neural network ensemble is able learn to play with good strategies in a complex, dynamic and difficult game environment which is not achievable by the individual neural network.

A robust framework for web information extraction and retrieval

The large volume of online and offline information that is available today has overwhelmed users’ efficiency and effectiveness in processing this information in order to extract relevant information. The exponential growth of the volume of Internet information complicates information access. Thus, it is a very time consuming and complex task for user in accessing relevant information. Information retrieval (IR) is a branch of artificial intelligence that tackles the problem of accessing and retrieving relevant information. The aim of IR is to enable the available data source to be queried for relevant information efficiently and effectively. This paper describes a robust information retrieval framework that can be used to retrieve relevant information. The proposed information retrieval framework is designed to assist users in accessing relevant information effectively and efficiently as it handles queries based on user preferences. Each component and module involved in the proposed framework will be explained in terms of functionality and the processes involved.