Ensuring sufficient liquidity is one of the key challenges for designers of prediction markets. Various market making algorithms have been proposed in the literature and deployed in practice, but there has been little effort to evaluate their benefits and disadvantages in a systematic manner. We introduce a novel experimental design for comparing market structures in live trading that ensures fair comparison between two different microstructures with the same trading population. Participants trade on outcomes related to a two-dimensional random walk that they observe on their computer screens. They can simultaneously trade in two markets, corresponding to the independent horizontal and vertical random walks. We use this experimental design to compare the popular inventory-based logarithmic market scoring rule (LMSR) market maker and a new information based Bayesian market maker (BMM). Our experiments reveal that BMM can offer significant benefits in terms of price stability and expected loss when controlling for liquidity; the caveat is that, unlike LMSR, BMM does not guarantee bounded loss. Our investigation also elucidates some general properties of market makers in prediction markets. In particular, there is an inherent tradeoff between adaptability to market shocks and convergence during market equilibrium.

In May 1, 2008, researchers at Hewlett Packard (HP) announced the first physical realization of a fundamental circuit element called memristor that attracted so much interest worldwide. This newly found element can easily be combined with crossbar interconnect technology which this new structure has opened a new field in designing configurable or programmable electronic systems. These systems in return can have applications in signal processing and artificial intelligence. In this paper, based on the simple memristor crossbar structure, we propose new and simple circuits for hardware implementation of fuzzy membership functions. In our proposed circuits, these fuzzy membership functions can have any shapes and resolutions. In addition, these circuits can be used as a basis in the construction of evolutionary systems.

- Memory refinements are designed below to detect those sequences of actions that have been repeated a given number n. Subsequently such sequences are permitted to run without CPU involvement. Actions are rehearsed and once learned, they are performed automatically without conscious involvement. Introduction This paper is written from the perspective of designing artificial brains that are modeled after human brains. This architectural work pays attention to human psychology including the interplay between short and long term memory, the realization of which is assumed constrained by circuit and system theory. This paper is a humble attempt, but at least it has an original aspect.

Low-rank matrix approximations are often used to help scale standard machine learning algorithms to large-scale problems. Recently, matrix coherence has been used to characterize the ability to extract global information from a subset of matrix entries in the context of these low-rank approximations and other sampling-based algorithms, e.g., matrix com- pletion, robust PCA. Since coherence is defined in terms of the singular vectors of a matrix and is expensive to compute, the practical significance of these results largely hinges on the following question: Can we efficiently and accurately estimate the coherence of a matrix? In this paper we address this question. We propose a novel algorithm for estimating coherence from a small number of columns, formally analyze its behavior, and derive a new coherence-based matrix approximation bound based on this analysis. We then present extensive experimental results on synthetic and real datasets that corroborate our worst-case theoretical analysis, yet provide strong support for the use of our proposed algorithm whenever low-rank approximation is being considered. Our algorithm efficiently and accurately estimates matrix coherence across a wide range of datasets, and these coherence estimates are excellent predictors of the effectiveness of sampling-based matrix approximation on a case-by-case basis.

We propose a simple kernel based nearest neighbor approach for handwritten digit classification. The "distance" here is actually a kernel defining the similarity between two images. We carefully study the effects of different number of neighbors and weight schemes and report the results. With only a few nearest neighbors (or most similar images) to vote, the test set error rate on MNIST database could reach about 1.5%-2.0%,

Ink Drop Spread (IDS) is the engine of Active Learning Method (ALM), which is the methodology of soft computing. IDS, as a pattern-based processing unit, extracts useful information from a system subjected to modeling. In spite of its excellent potential in solving problems such as classification and modeling compared to other soft computing tools, finding its simple and fast hardware implementation is still a challenge. This paper describes a new hardware implementation of IDS method based on the memristor crossbar structure. In addition of simplicity, being completely real-time, having low latency and the ability to continue working after the occurrence of power breakdown are some of the advantages of our proposed circuit.

We formulate weighted graph clustering as a prediction problem: given a subset of edge weights we analyze the ability of graph clustering to predict the remaining edge weights. This formulation enables practical and theoretical comparison of different approaches to graph clustering as well as comparison of graph clustering with other possible ways to model the graph. We adapt the PAC-Bayesian analysis of co-clustering (Seldin and Tishby, 2008; Seldin, 2009) to derive a PAC-Bayesian generalization bound for graph clustering. The bound shows that graph clustering should optimize a trade-off between empirical data fit and the mutual information that clusters preserve on the graph nodes. A similar trade-off derived from information-theoretic considerations was already shown to produce state-of-the-art results in practice (Slonim et al., 2005; Yom-Tov and Slonim, 2009). This paper supports the empirical evidence by providing a better theoretical foundation, suggesting formal generalization guarantees, and offering a more accurate way to deal with finite sample issues. We derive a bound minimization algorithm and show that it provides good results in real-life problems and that the derived PAC-Bayesian bound is reasonably tight.

In this paper we introduce a novel method for automatically tuning the search parameters of a chess program using genetic algorithms. Our results show that a large set of parameter values can be learned automatically, such that the resulting performance is comparable with that of manually tuned parameters of top tournament-playing chess programs.

Classic evaluation methods of believable agents are time-consuming because they involve many human to judge agents. They are well suited to validate work on new believable behaviours models. However, during the implementation, numerous experiments can help to improve agents' believability. We propose a method which aim at assessing how much an agent's behaviour looks like humans' behaviours. By representing behaviours with vectors, we can store data computed for humans and then evaluate as many agents as needed without further need of humans. We present a test experiment which shows that even a simple evaluation following our method can reveal differences between quite believable agents and humans. This method seems promising although, as shown in our experiment, results' analysis can be difficult.

ABSTRACT In this paper, we address the problem of creating believable agents (virtual characters) in video games. We consider only one meaning of believability, "giving the feeling of being controlled by a player", and outline the problem of its evaluation. We present several models for agents in games which can produce believable behaviours, both from industry and research. For high level of believability, learning and especially imitation learning seems to be the way to go. We make a quick overview of different approaches to make video games' agents learn from players. To conclude we propose a two-step method to develop new models for believable agents. First we must find the criteria for believability for our application and define an evaluation method. Then the model and the learning algorithm can be designed. INTRODUCTION Nowadays, more and more consoles and video games are designed to make the player feel like he/she is in the game. To define how well this goal is achieved, two criteria have been defined in academic research: immersion and presence. According to Slater, immersion is an objective criterion which depends on the hardware and software(Slater et al. 1995). It includes criteria based on virtual sensory information's types, variety, richness, direction and in which extend they override real ones. For example, force feedback and motion sensing controllers, surround sound and high dynamic range rendering can improve the immersion. Presence, also known as telepresence (Steuer 1992), is a more subjective criterion.