Abstract: The Kushilevitz-Mansour (KM) algorithm is an algorithm that finds all the "large" Fourier coefficients of a Boolean function. It is the main tool for learning decision trees and DNF expressions in the PAC model with respect to the uniform distribution. The algorithm requires access to the membership query (MQ) oracle. The access is often unavailable in learning applications and thus the KM algorithm cannot be used. We significantly weaken this requirement by producing an analogue of the KM algorithm that uses extended statistical queries (SQ) (SQs in which the expectation is taken with respect to a distribution given by a learning algorithm). We restrict a set of distributions that a learning algorithm may use for its statistical queries to be a set of product distributions with each bit being 1 with probability ρ, 1/2 or 1-ρ for a constant 1/2 > ρ > 0 (we denote the resulting model by SQ-Dρ). Our analogue finds all the "large" Fourier coefficients of degree lower than clog(n) (we call it the Bounded Sieve (BS)). We use BS to learn decision trees and by adapting Freund's boosting technique we give an algorithm that learns DNF in SQ-Dρ. An important property of the model is that its algorithms can be simulated by MQs with persistent noise. With some modifications BS can also be simulated by MQs with product attribute noise (i.e., for a query x oracle changes every bit of x with some constant probability and calculates the value of the target function at the resulting point) and classification noise. This implies learnability of decision trees and weak learnability of DNF with this non-trivial noise. In the second part of this paper we develop a characterization for learnability with these extended statistical queries. We show that our characterization when applied to SQ-Dρ is tight in terms of learning parity functions. We extend the result given by Blum et al. by proving that there is a class learnable in the PAC model with random classification noise and not learnable in SQ-Dρ.

Abstract: Carin-ALN is an interesting new rule learning bias for ILP By allowing description logic terms as predicates of literals in datalog rules, it extends the normal bias used in ILP as it allows the use of all quantified variables in the body of a clause It also has at-least and at-most restrictions to access the amount of indeterminism of relations From a complexity point of view Carin-ALN allows to handle the difficult indeterminate relations efficiently by abstracting them into determinate aggregations This paper describes a method which enables the embedding of Carin-ALN rule subsumption and learning into datalog rule subsumption and learning with numerical constraints On the theoretical side, this allows us to transfer the learnability results known for ILP to Carin-ALN rules On the practical side, this gives us a preprocessing method, which enables ILP systems to learn Carin-ALN rules just by transforming the data to be analyzed We show, that this is not only a theoretical result in a first experiment: learning Carin-ALN rules from a standard ILP dataset

Abstract: The benefits of using Model-Based Reasoning for domain modeling are several-fold. We analyze these benefits and illustrate them in the context of a Web-based Intelligent Tutoring System. The system is designed to teach students to analyze and debug C++ programs for semantic and run-time errors. We have evaluated one instance of the Model-Based tutor, which deals with debugging pointers in C++, in several sections of Computer Science II course. We will present the results of these evaluations, which confirm the learnability of Model-Based tutors.

Abstract: This paper is concerned with problems of learnability. It speculates on how learners, with the help of UG and a few common-sense strategies, go about discovering abstract relationships between superficially different structural formats available in the input. Far from being confused by variation, learners can use what they perceive as conflicts between UG and experience to infer new system properties. According to the multiple-roots perspective proposed here, the monolingual child starts out like a bilingual child, that is, with coexisting (but not arbitrary) sentential roots, eventually deciding where convergence is possible and where (as in the case of the real bilingual) it is not. The knowledge domain for which this scenario is explored is the acquisition of finite and nonfinite verb placement in German. The paper also addresses the issue of how different target languages enhance or slow down the overall process of structure building and relates this to asynchronies observed in bilingual children.

Abstract: We present an account of human concept learning—that is, learning of categories from examples—based on the principle of minimum description length (MDL). In support of this theory, we tested a wide range of two-dimensional concept types, including both regular (simple) and highly irregular (complex) structures, and found the MDL theory to give a good account of subjects' performance. This suggests that the intrinsic complexity of a concept (that is, its description length) systematically influences its learnability.

Abstract: We prove a new combinatorial characterization of polynomial learnability from equivalence queries, and state some of its consequences relating the learnability of a class with the learnability via equivalence and membership queries of its subclasses obtained by restricting the instance space. Then we propose and study two models of query learning in which there is a probability distribution on the instance space, both as an application of the tools developed from the combinatorial characterization and as models of independent interest.

Abstract: A previous report identified several usability and learnability problems with integrated development environments (IDE) for Java. That report also cast these problems as examples of a conceptual gap between developer mental models and how programs are represented in IDEs. This present study extends the previous work through heuristic and psychometric assessment of problems reported by both experienced and inexperienced developers in their use of an IDE for C++. The results indicate that both groups identified similar kinds of ease-of-use problems, especially concerning program learnability and visibility (e.g., the usefulness of error and help messages). These findings are discussed in relation to other research results about developers' experiences with CASE tools and conceptual gaps between the tools and their users.

Abstract: This paper presents a study on associative mental arithmetic with mean-field Boltzmann Machines. We examined the role of number representations, showing theoretically and experimentally that cardinal number representations (e.g., numerosity) are superior to symbolic and ordinal representations w.r.t. learnability and cognitive plausibility. Only the network trained on numerosities exhibited the problem-size effect, the core phenomenon in human behavioral studies. These results urge a reevaluation of current cognitive models of mental arithmetic.

Abstract: This paper studies α-CoAgnostic learnability of classes of boolean formulas. To α-CoAgnostic learn C from H, the learner seeks a hypothesis h ∈ H that agrees (rather than disagrees as in Agnostic learning) within a factor a of the best agreement of any f E C. Although 1-CoAgnostic learning is equivalent to Agnostic learning, this is not true for α-CoAgnostic learning for < α < 1. It is known that α-CoAgnostic learning algorithms are equivalent to α-approximation algorithms for maximum agreement problems. Many studies have been done on maximum agreement problems, for classes such as monomials, monotone monomials, antimonotone monomials, halfspaces and balls. We study these problems further and some extensions of them. For the above classes we improve the best previously known factors a for the hardness of α-CoAgnostic learning. We also find the first constant lower bounds for decision lists, exclusive-or, halfspaces (over the boolean domain), 2-term DNF and 2-term multivariate polynomials.

Abstract: We prove upper bounds for combinatorial parameters of finite relational structures, related to the complexity of learning a definable set. We show that monadic second order (MSO) formulas with parameters have bounded VC-dimension over structures of bounded clique-width, and first-order formulas with parameters have bounded VC-dimension over structures of bounded local clique-width (this includes planar graphs). We also show that MSO formulas of a fixed size have bounded strong consistency dimension over MSO formulas of a fixed larger size, for colored trees. These bound simply positive learnability results for the PAC and equivalence query learnability of a definable set over these structures. The proofs are based on bounds for related definability problems for tree automata.

Abstract: How can an intelligent agent learn an effective representation of its world? This dissertation applies the psychological principle of cognitive economy to the problem of representation in reinforcement learning. Psychologists have shown that humans cope with difficult tasks by simplifying the task domain, focusing on relevant features and generalizing over states of the world which are “the same” with respect to the task. This dissertation defines a principled set of requirements for representations in reinforcement learning, by applying these principles of cognitive economy to the agent's need to choose the correct actions in its task. The dissertation formalizes the principle of cognitive economy into algorithmic criteria for feature extraction in reinforcement learning. To do this, it develops mathematical definitions of feature importance, sound decisions, state compatibility, and necessary distinctions, in terms of the rewards expected by the agent in the task. The analysis shows how the representation determines the apparent values of the agent's actions, and proves that the state compatibility criteria presented here result in representations which satisfy a criterion for task learnability. The dissertation reports on experiments that illustrate one implementation of these ideas in a system which constructs its representation as it goes about learning the task. Results with the puck-on-a-hill task and the pole-balancing task show that the ideas are sound and can be of practical benefit. The principal contributions of this dissertation are a new framework for thinking about feature extraction in terms of cognitive economy, and a demonstration of the effectiveness of an algorithm based on this new framework.

Abstract: In this article I present the concept of ‘focus on learnable form’ and show how it could be implemented in the classroom. ‘Focus on form’ research has produced increasing evidence that a form focus can improve the acquisition of the particular form while remaining compatible with the communicative approach. The learnability of grammatical form, a key issue in this research, has been addressed in some studies by identifying the emergence of the form. Since there are problems in relying on emergence, I argue for the adoption of Processability Theory (Pienemann, 1998) and specifically the construct of developmental stages. This framework enables teachers to predict the forms that would be beneficial to focus on and those that are developmentally too advanced for effective focus on form. Despite criticisms of ‘structural’ approaches to SLA research, Processability Theory has a lot to offer communicative language teaching. As I have found in teaching ESL to adolescent and adult learners, ‘focus on learnable form’ in a communicative context is achievable in the classroom and can be implemented as one component of a communicative curriculum by following three steps: assessing the learners, selecting a ‘learnable’ form and focusing on this learnable form in a communicative context.

Abstract: Two major issues in formal language learnability are the problem of learning restrictive distributions (sometimes known as the “subset problem”), and the problem of structural ambiguity. While substantial progress has been made in addressing each of these problems in isolation, a complication can arise when a learner is faced with a learning situation that exhibits both problems. It is possible for the two problems to interact: allowing grammars of differing restrictiveness can complicate efforts to contend with structural ambiguity. The main result of this paper is a demonstration that a construction already proposed for learning restrictive grammars, the r-measure, can be used to contend with the complications in structural ambiguity that result from the existence of grammars of differing restrictiveness.

Abstract: INTRODUCTION In recent years, tangible user interaction (TUI) has attracted attention among researchers, who have come to see the screen, keyboard, and mouse interaction with present computers as limiting. Although humans have developed sophisticated skills in the recognition and manipulation of objects over thousands of years, the displays and push buttons of today's electronic products far from exploit these skills. Current electronic products emphasise sequences of digital inputs that mirror the underlying software control structure, while exploiting only a very limited range of cognitive and motoric skills. Although some interfaces of this genre are simple to use, many are limiting and frustrating. 'Skill' in using electronic products is more about executing fixed recipes than about honing of our motoric ability (Maeda, 1999). Yet there is evidence in the literature that the physical use of objects improves the learning process (Brereton 1998). If we assume that by using TUI we can leverage the user's motoric ability to improve the learnability, usability and experience of interactive products, this leads to questions such as: • How do motoric and cognitive skills develop as we use and manipulate tools and interfaces? • What makes actions scalable along the learning curve, i.e. which actions allow us to improve our skills through practice to become 'pros' without creating a threshold which excludes novices? • What makes physical interaction an interesting, beautiful or fun experience, thus enticing users to further polish their skills? * How do we successfully map actions to functionality? TUI design is still a young field, and the term is used to cover a broad range of design approaches. The purpose of this research workshop is to bring together researchers to explore relationships between TUI, learning and functionality, and--if possible--to establish a framework for TUI approaches. PROGRAM 1, Introduction to physical interaction and learning 2. Designs; Show & tell Participants demonstrate their own example of tangible user inter- face design, either 'live' or by means of video. Grouping of approaches to tangible interaction design. 3. Gestures" Act & imitate In groups of four, participants present an object they have brought along and that they are expert in handling. A novice tries to imitate. Both expert and novice are recorded on video. 4, Matrix of skills Teams present gestures and discuss whether these are learnable, scalable along the learning curve, interesting, beautiful, fun etc. Video images are grouped and projected as a 'matrix of skills', which runs all day (Buur & Sendergaard 2000). 5. Design & tinkering Based on favourite gestures and design approaches, the teams work on a small design assignment, using a variety of tinkering materials provided by the organisers (Hummels et al., 2001). Teams act out product use, using gestures and props. 6. Group discussion • What does tangibility mean for learning and vice versa? • How do we improve the fit between actions and functions? • Can we categorise approaches in tangible interaction? OUTCOME The result of the workshop is presented as a videoposter at the conference. It shows excerpts of both novice and pro gestures, grouped and annotated as they were during the workshop by the participants. REFERENCES 1. Brereton, M.E (1999) The Role of Hardware in Learning Engineering Fundamentals; an Empirical Study of Engineering Design and Product Analysis Activity. PhD Dissertation, Stanford University. 2. Buur, J and Sondergaard, A. (2000): Video Card Game: An Augmented Environment for User Centred Design Discussions. In Proceedings of DARE 2000, Designing Augmented Reality Environments, ACM, Elsinore. 3. Hummels, C.C.M., Djajadiningrat, J.P., & Overbeeke, C.J. (2001). Knowing, doing and feeling: communicating with your digital products. Interdisziplin&res Kolleg Kognitions- und Neurowissenschaften, G0nne am M6hnesee, March 2- 9 2001, 289-308. 4. Maeda, J., & Antonelli, P. (1999). Design by Numbers. MIT Press. DIS2002, London © Copyright 2002 ACM 1-58113-2-9-0/00/0008

Abstract: The central topic of the paper is the learnability of the recursively enumerable subspaces of V∞/V, where V∞ is the standard recursive vector space over the rationals with countably infinite dimension, and V is a given recursively enumerable subspace of V∞. It is shown that certain types of vector spaces can be characterized in terms of learnability properties: V∞/V is behaviourally correct learnable from text iff V is finitely dimensional, V∞/V is behaviourally correct learnable from switching type of information iff V is finite-dimensional, 0-thin, or 1-thin. On the other hand, learnability from an informant does not correspond to similar algebraic properties of a given space. There are 0-thin spaces W 1 and W 2 such that W 1 is not explanatorily learnable from informant and the infinite product (W 1 )∞ is not behaviourally correct learnable, while W 2 and the infinite product (W 2 )∞ are both explanatorily learnable from informant.

Abstract: This paper presents a model of the origins of syllable systems that brings plausibility to the theory which claims that language learning, and in particular phonological acquisition, needs not innate linguistically specific information, as believed by many researchers of the Chomskyan school, but is rather made possible by the interaction between general motor, perceptual, cognitive and social constraints through a self-organizing process. The strategy is to replace the question of acquisition in a larger and evolutionary (cultural) framework: the model addresses the question of the origins of syllable systems (syllables are the major phonological units in speech). It is based on the artificial life methodology of building a society of agents, endowed with motor, perceptual and cognitive apparati that are generic and realistic. We show that agents effectively build sound systems and how these sound systems relate to existing human sound systems. Results concerning the learnability of the produced sound systems by fresh/baby agents are detailed: the critical period effect and the artificial language effect can effectively be predicted by our model. The ability of children to learn sound systems is explained by the evolutionary history of these sound systems, which were precisely shaped so as to fit the ecological niche formed by the brains and bodies of these children, and not the other way around (as advocated by Chomskyan approaches to language).

Abstract: The design, development and formative evaluation of an Electronic Portfolio Template System for Cycle 1 students in the Quebec Education System is described. The prototype is a web-based, database-driven process and showcase portfolio container that facilitates portfolio development. This system contains administrator, teacher and student environments. Each of these environments, along with the installation, set-up and documentation process was evaluated. In all, twenty-six participants evaluated the various environments and processes. Results of all evaluations are presented. The student environment received the most feedback with strengths reported relating to interface design, usability, learnability and aesthetics and weaknesses reported relating to suitability and navigation. Interface design, learnability and aesthetics were reported as strengths while marginal navigation weaknesses were reported in the teacher and administrator environments. Evaluative comments, recommendations for improvement and suggestions for further research are presented.

Abstract: Accessibility of legal sources is crucial to both jurists and citizens. As the development of E-government is speeded up the number of legal information systems including knowledge-based services is likely to increase accordingly. In the Program for an Ontology-based Working Environment for Rules and legislation (POWER) the Dutch Tax and Customs Administration (Belastingdienst) developed a formal model- ling approach for modelling legal sources. The results of applying this modelling pro- cess are formal models expressed in UML/OCL which can be used as the basis for amongst others verification, simulation and application generation. Since the legal knowledge to be applied in the operational units and on the Web is represented in these formal models, their quality needs serious attention. To be able to determine their quality, inspection by (legal) experts should be possible (validation). In most cases however these experts are not able to read these UML/OCL-models. Therefore a representation is required that can easily be derived from these models and that is easy to understand. Based on previous research we compared two different knowledge representations and tested their learnability and usability. Furthermore we examined the effect of rep- resentation on legal decision-making. The two representation-forms used are produc- tion rules and scenarios. The results of our experiment show that the performance of scenarios is significantly better then the performance of production rules.

Abstract: The central topic of the paper is the learnability of the recursively enumerable subspaces of V?/V, where V? is the standard recursive vector space over the rationals with countably infinite dimension, and V is a given recursively enumerable subspace of V?. It is shown that certain types of vector spaces can be characterized in terms of learnability properties: V?/V is behaviourally correct learnable from text iff V is finitely dimensional, V?/V is behaviourally correct learnable from switching type of information iff V is finite-dimensional, 0-thin, or 1-thin. On the other hand, learnability from an informant does not correspond to similar algebraic properties of a given space. There are 0-thin spaces W1 and W2 such that W1 is not explanatorily learnable from informant and the infinite product (W1)? is not behaviourally correct learnable, while W2 and the infinite product (W2)? are both explanatorily learnable from informant.

Abstract: We give an algorithm for learning a more expressive circuit class than the class AC/sup 0/ considered by Linial et al. (1993) and Kharitonov (1993). The new algorithm learns constant-depth AND/OR/NOT circuits augmented with (a limited number of) majority gates. Our main positive result for these circuits is stated informally.

Abstract: The learnability of a software user interface is an important aspect of software design. This study evaluates the learnability of an instructional game entitled Justice Bao. It attests issues related to "button-less" interface that uses an object on screen as a "clickable" object for various functions. It also highlights some issues with regard to the use of text labels to instruct users on the functions of a "clickable" object. Some guidelines in the design of button-less interface are suggested. ********** The interface of instructional multimedia plays an important role in framing students' learning experience because it instructs the learner how to behave within the created environment (Lohr, 2000). It delimitates its use and thus the user experience (Mardsjo, 1996). The interface has significant impacts on the way the user thinks and works (Rheingold, 1990). Thus, researchers (Kuittinen, 1998; Pham, 1998) have suggested that the evaluation of user interface is one of the important components of the research and development cycle of instructional software. This study evaluated the learnability of the interface of an instructional game entitled Justice Bao, designed to motivate students in doing Chinese language drill-and-practice questions at the primary school level. Learnability is defined as the ease with which new or occasional users may accomplish some task using the interface. It is commonly measured by the number of trails, that a new user needs to complete a task without being trained (Lingaard, 1994). The motivation for evaluating the learnability of the Justice Bao interface is threefold. First, as instructional software, its interface should not be difficult to learn and not interfere with the learning of the subject matter (Lee, 1996). Second, as a game, its interface design should be easy to learn or the player may abandon it quickly (Crawford, 1990). Third and most importantly, the interface of Justice Bao adopted an experimental design format, the "button-less" interface. We are particularly interested in exploring how users perceive this button-less format. Chen, Wong, and Hsu (2002) posited that the interface design of instructional software should use objects on screen, as much as possible, as "clickable" objects to activate various functions and interactions. They argued that the usual design of buttons and icons are metaphorically inappropriate and objects on the screen that are more contextually meaningful could replace them. Consequently, the buttonless design format may yield a more learnabl e interface. The present evaluation attests issues related to the "button-less" interface. Findings of this study will advance our understanding of learnability of the button-less interface design format and provide practical suggestions for designing more usable interface. The Justice Bao Instructional Game Justice Bao is divided into two inter-related components: (a) the study of the case that is carried out in the courtroom (Figure 1) and (b) the investigation process that is carried out in the town (Figure 2). The story opens with an animation of a crime in progress. A thief has stolen some valuable that was hidden in a temple. A monk discovered the crime and brought it before Justice Bao. Justice Bao then summoned the three suspects to the courtroom to give their statements. The three suspects are the monk, a carpenter, and a scholar. In the courtroom, the user plays the role of Justice Bao. He or she activates the statements by clicking the suspects in turns. After listening to the suspects' statements, the user sends a detective into the town. In the town, the user assumes the role of the detective. To solve the case, the detective navigates through the different areas of the town, interacts with the people to gather information and answers numerous questions posed by the system. …

Abstract: Elements of skill-development in using in-vehicle-information-systems are discussed, followed by results of an on-the-road study. Entering of a destination into a route guidance systems was practised by 12 drivers while driving on routes with reduced traffic. Each driver performed 68 to 100 training trials. Two route guidance systems were used which differed in user interface. Mean duration of destination entry and mean frequency of glances to the in-car task diminished with practice and varied with difficulty of route. Individual differences in performance and skill acquisition were also found, while task demands remain high after practice. Differences between user interfaces are apparent in performance.

Abstract: The paper develops the theory of refutable/inductive learning as a foundation of discovery science from examples. We consider refutable/inductive language learning from positive examples, some of which may be incorrect. The error or incorrectness we consider is the one described uniformly in terms of a distance over strings. We define a k-neighbor closure of a language L as the collection of strings each of which is at most k distant from some string in L. In ordinary learning paradigm, a target language is assumed to belong to a hypothesis space without any guarantee. In this paper, we allow an inference machine to infer a neighbor closure instead of the original language as an admissible approximation. We formalize such kind of learning, and give some sufficient conditions for a hypothesis space.As its application to concrete problems, we deal with languages defined by decision trees over patterns. The problem of learning decision trees over patterns has been studied from a viewpoint of knowledge discovery for Genome information processing in the framework of PAC learning from both positive and negative examples. We investigate their learnability in the limit from neighbor examples as well as refutable learnability from complete examples, i.e., from both positive and negative examples. Furthermore, we present some procedures which plays an important role for designing efficient learning algorithms for decision trees over regular patterns.