Abstract
A trading strategy is an algorithm that provides decision support for a trader. An ideal system suggests which stocks to buy and sell at every moment. Limited but still very useful trading strategies suggest stocks to buy, but leave the sell decisions and the decision of proportions of different stocks to the trader, or to another automatic decision mechanisim. In this paper we use a previously introduced method of predicting rank variables to produce both buy and sell decisions. The rank variables are predicted by neural networks, and provide an efficient way to produce daily buy (and also sell) suggestions. This should be seen in contrast to "ordinary" technical indicators that often give very few signals, or buy/sell signals for many stocks at the same time. The produced buy signals are further processed in a classification module that aims at identifying which of the numerous buy signals one should trust, and of which ones one should discard. The classification reduces the number of buy signals and also increases both hitrate and overall profit for a simulated trader. Data from the US stock market for 1992-2001 is used in the tests of the system, and the results show how a trading system's performance can be significantly improved by adding a post-processing classification layer between the generation of trading signals and the actual decision making.

Abstract
Hyperbolic invariant sets A of C1+gamma diffeomorphisms where either the stable or unstable leaves are 1-dimensional are considered in this paper, Under the assumption that the A has local product structure, the authors prove that the holonomies between the 1-dimensional leaves are C1+alpha for some 0 < alpha < 1.

Abstract
There is currently an increasing interest in exploring the opportunities for competitive advantage that can be gained by reinforcing core competencies and innovative capabilities through networks of industrial and business partners. This paper firstly identifies some of the gaps that exist within current information systems that claim to support eBusiness and eWork in networked enterprises and describes some of the general requirements of distributed and decentralised information systems for companies operating in networks. It goes on to cover some principles for the design of a distributed IS providing an advanced infrastructure to support general co-operation, particular methodologies for co-operative and collaborative planning and guidelines for network set-up and support. The present work is one of the areas currently being delivered as part of the Europe-an IST consortium called Co-OPERATE. A distributed and decentralised information system, based on an architecture of agents and extensively using the internet, is being designed and implemented as a means to provide new and more powerful decision support tools for networked enterprises.

Abstract
We construct a Teichmuller space for the C1+-conjugacy classes of hyperbolic dynamical systems on surfaces. After introducing the notion of an HR structure which associates an affine structure with each of the stable and unstable laminations, we show that there is a one-to-one correspondence between these HR structures and the C1+-conjugacy classes. As part of the proof we construct a canonical representative dynamical system for each HR structure. This has the smoothest holonomies of any representative of the corresponding C1+-conjugacy class. Finally, we introduce solenoid functions and show that they provide a good Teichmuller space.

Abstract

Abstract
Several researchers have studied the application of Machine Learning techniques to the task of user modeling. As most of them pointed out, this task requires learning algorithms that should work on-line, incorporate new information incrementality, and should exhibit the capacity to deal with concept-drift. In this paper we present Adaptive Bayes, an extension to the well-known naive-Bayes, one of the most common used learning algorithms for the task of user modeling. Adaptive Bayes is an incremental learning algorithm that could work on-line. We have evaluated Adaptive Bayes on both frameworks. Using a set of benchmark problems from the UCI repository [2], and using several evaluation statistics, all the adaptive systems show significant advantages in comparison against their non-adaptive versions.