We provide an algorithm to compute near-optimal strategies for the Cultaptation social learning game. We show that the strategies produced by our algorithm are near-optimal, both in their expected utility and their expected reproductive success. We show how our algorithm can be used to provide insight into evolutionary conditions under which learning is best done by copying others, versus the conditions under which learning is best done by trial-and-error.

This survey is an introduction to positive definite kernels and the set of methods they have inspired in the machine learning literature, namely kernel methods. We first discuss some properties of positive definite kernels as well as reproducing kernel Hibert spaces, the natural extension of the set of functions $\{k(x,\cdot),x\in\mathcal{X}\}$ associated with a kernel $k$ defined on a space $\mathcal{X}$. We discuss at length the construction of kernel functions that take advantage of well-known statistical models. We provide an overview of numerous data-analysis methods which take advantage of reproducing kernel Hilbert spaces and discuss the idea of combining several kernels to improve the performance on certain tasks. We also provide a short cookbook of different kernels which are particularly useful for certain data-types such as images, graphs or speech segments.

Much recent work in bioinformatics has focused on the inference of various types of biological networks, representing gene regulation, metabolic processes, protein-protein interactions, etc. A common setting involves inferring network edges in a supervised fashion from a set of high-confidence edges, possibly characterized by multiple, heterogeneous data sets (protein sequence, gene expression, etc.). Here, we distinguish between two modes of inference in this setting: direct inference based upon similarities between nodes joined by an edge, and indirect inference based upon similarities between one pair of nodes and another pair of nodes. We propose a supervised approach for the direct case by translating it into a distance metric learning problem. A relaxation of the resulting convex optimization problem leads to the support vector machine (SVM) algorithm with a particular kernel for pairs, which we call the metric learning pairwise kernel (MLPK). We demonstrate, using several real biological networks, that this direct approach often improves upon the state-of-the-art SVM for indirect inference with the tensor product pairwise kernel.

Preface This is a companion paper to Conscious Intelligent Systems Part 1 by the same author (1), which discusses a possible evolutionary path for consciousness and intelligence from simple systems to human level consciousness and intelligence. Man has long been held to be a thinking animal, his thought processes have been held to be the reason for his superiority over the animals. The grand aim of AI has always been to make an entity that can think. Turing took up this very question in his paper (2) on whether machines can think. On the more prosaic roads that real AI has been forced to follow, such grand questions have almost died down. Another major trigger for the demise has been Searle's Chinese Room (3) parody . With this rather cunning device, Searle set the cat among the pigeons and has helped induce self-doubt in the best of AI theorists. One of the major triggers towards Searle's views was language, whether syntax suffices for semantics and therefore understanding. From our evolutionary learning system perspective, which we discuss in Part I of this discussion, we see that all these processes are tied together, the processes of consciousness, intelligence, mind, thought, and language. In a bid to show the interconnectedness of these factors, we take up the question of understanding and its communication. Similar to our treatment of the subject of consciousness based intelligent systems in Part 1, here we treat understanding from first principles. Understanding In the real world when we use the term understanding, it has two main attributes; one is the capacity to infer, the other is the capacity to recognize or discern. In computing and AI contexts the word understanding is arguably tilted more in favor of inference than perception or cognition, in normal life and in the natural kingdom the reverse is true. This is primarily because AI's aims and present status look elemental when compared to the entities of the natural world. The other reason is that AI entities find it easier to infer than cognize, which is in itself a reflection of their design sources and its aims. For the purposes of this discussion the term understanding implies the natural version, a mix of cognition and inference. If we start from first principles, it is clear that for a rule to emerge out of a set of raw data, an inferential process has to run on it. This process could be a formal inferential process or a process that is driven by the needs of economy or efficiency. Rules need not always rise out of intentional activity, for instance the interaction of water flowing from an open tap into a pot already full of water can create a set of rules that disallow further water entry, limit mixing and regulate overflow, many natural rules rise from interactions like these.

Did natural consciousness and intelligent systems arise out of a path that was co-evolutionary to evolution? Can we explain human self-consciousness as having risen out of such an evolutionary path? If so how could it have been? In this first part of a two-part paper (titled IXI), we take a learning system perspective to the problem of consciousness and intelligent systems, an approach that may look unseasonable in this age of fMRI's and high tech neuroscience. We posit conscious intelligent systems in natural environments and wonder how natural factors influence their design paths. Such a perspective allows us to explain seamlessly a variety of natural factors, factors ranging from the rise and presence of the human mind, man's sense of I, his self-consciousness and his looping thought processes to factors like reproduction, incubation, extinction, sleep, the richness of natural behavior, etc. It even allows us to speculate on a possible human evolution scenario and other natural phenomena.

Latent Semantic Analysis (LSA) is a widely used Information Retrieval method based on "bag-of-words" assumption. However, according to general conception, syntax plays a role in representing meaning of sentences. Thus, enhancing LSA with part-of-speech (POS) information to capture the context of word occurrences appears to be theoretically feasible extension. The approach is tested empirically on a automatic essay grading system using LSA for document similarity comparisons. A comparison on several POS-enhanced LSA models is reported. Our findings show that the addition of contextual information in the form of POS tags can raise the accuracy of the LSA-based scoring models up to 10.77 per cent.

This paper is based on my invited talk at the 19th Annual Conference on Learning Theory (Pittsburgh, PA, June 24, 2006). In recent years COL T invited talks have tended to aim at establishing connections between the traditio nal concerns of the learning community and the work done by other communities (s uch as game theory, statistics, information theory, and optimization). F ollowing this tradition, I will argue that some ideas from the foundations of prob ability can be fruitfully applied in competitive on-line learning. In this paper I will use the following informal taxonomy of predictions (reminiscent of Shafer's [36], Figure 2, taxonomy of probabilities): D-predictions are mere Decisions. They can never be true or false but can be good or bad.

E-learning is nowadays one of the most interesting of the "e- " domains available through the Internet. The main problem to create a Web-based, virtual environment is to model the traditional domain and to implement the model using the most suitable technologies. We analyzed the distance learning domain and investigated the possibility to implement some e-learning services using mobile agent technologies. This paper presents a model of the Student Assessment Service (SAS) and an agent-based framework developed to be used for implementing specific applications. A specific Student Assessment application that relies on the framework was developed.

The Unified Software Development Process (USDP) and UML have been now generally accepted as the standard methodology and modeling language for developing Object-Oriented Systems. Although Agent-based Systems introduces new issues, we consider that USDP and UML can be used in an extended manner for modeling Agent-based Systems. The paper presents a methodology for designing agent-based systems and the specific models expressed in an UML-based notation corresponding to each phase of the software development process. UML was extended using the provided mechanism: stereotypes. Therefore, this approach can be managed with any CASE tool supporting UML. A Case Study, the development of a specific agent-based Student Evaluation System (SAS), is presented.

A fundamental problem in artificial intelligence is that nobody really knows what intelligence is. The problem is especially acute when we need to consider artificial systems which are significantly different to humans. In this paper we approach this problem in the following way: We take a number of well known informal definitions of human intelligence that have been given by experts, and extract their essential features. These are then mathematically formalised to produce a general measure of intelligence for arbitrary machines. We believe that this measure formally captures the concept of machine intelligence in the broadest reasonable sense.