This paper studies the problem of ergodicity of transition probability matrices in Markovian models, such as hidden Markov models (HMMs), and how it makes very difficult the task of learning to represent long-term context for sequential data. This phenomenon hurts the forward propagation of long-term context information, as well as learning a hidden state representation to represent long-term context, which depends on propagating credit information backwards in time. Using results from Markov chain theory, we show that this problem of diffusion of context and credit is reduced when the transition probabilities approach 0 or 1, i.e., the transition probability matrices are sparse and the model essentially deterministic. The results found in this paper apply to learning approaches based on continuous optimization, such as gradient descent and the Baum-Welch algorithm.

The first session aimed to set the component being tested. The stage for the day's discussion by focusing variation in all three of these contexts on the issues surrounding the will lead to different types of and Verification of Knowledge-use of formal specification techniques The first paper, by Formal Specifications to Design Intelligence (AAAI-94) in Seattle, Lance Miller of SAIC, was entitled Verifiable Hybrid KBS" by Rose Gamble, Washington, marked the seventh This paper provided a with its specification, and (2) the The 1994 workshop was significant basis for the comparison of validation refinement of formal specifications in that there was a definitive move in and verification techniques to for their implementation. O'Leary, from the lows the possibility of constraining techniques for validating certain University of Southern California, the experts' choices to ensure that properties of KBSs. A paper by presented a paper on the relationship any new knowledge added is valid Alun Preece, Cliff Gossner, and T. between errors and size in KBSs. This and that the knowledge base structure Radhakrishnan (all from the University paper is among the first to address ensures the knowledge is of Aberdeen, Scotland) considered this important issue.

Analogy has a rich history in Western civilization. Over the centuries, it has become reified in that analogical reasoning has sometimes been regarded as a fundamental cognitive process. In addition, it has become identified with a particular expressive format. The limitations of the modern view are illustrated by monster analogies, which show that analogy need not be regarded as something having a single form, format, or semantics. Analogy clearly does depend on the human ability to create and use well-defined or analytic formats for laying out propositions that express or imply meanings and perceptions. Beyond this dependence, research in cognitive science suggests that analogy relies on a number of genuinely fundamental cognitive capabilities, including semantic flexibility, the perception of resemblances and of distinctions, imagination, and metaphor. Extant symbolic models of analogical reasoning have various sorts of limitation, yet each model presents some important insights and plausible mechanisms. I argue that future efforts could be aimed at integration. This aim would include the incorporation of contextual information, the construction of semantic bases that are dynamic and knowledge rich, and the incorporation of multiple approaches to the problems of inference constraint.

Prizes were awarded to students to encourage improved research. Michel Wermelinger, Universidade Nova de Lisboa, Portugal, was the winner of the best paper award for his work "Basic Conceptual Structure Theory," which provided a significant In "Representations Technology, Bangkok, Thailand, won Papers were presented by a number interest in the use of conceptual he Second International Conference (ICCS'94) was held at the of individuals and groups from graphs. The funds were made available University of Maryland, College several countries on the development through a grant from the American Park, Maryland, on August 16 to 20. and use of the conceptual Association for Artificial Intelligence The conference marked the tenth graph representational language. Sponsors included the University of Graph Workbench," chaired by Gerard vice-president of academic affairs, Paradigm Development Corp. in Urbana, Illinois, was the second She received her Ph.D. from the introduction, "Aristotelian and

Several of the workshops have led to discourse? At what levels of the art in the field (Dale et al. 1992; generation is information processed on Natural Language Paris, Swartout, and Mann 1991; How can we generate to 24 June 1994 at the Nonantum The goal of this latest workshop multilingual texts efficiently? Inn on the seacoast in Kennebunkport, was to introduce new, cutting-edge The topics presented at the workshop Maine. Two invited speakers described subtopics such as evaluation, casual site contributed greatly to their perspectives on two areas outside explanation generation, and summarization the success of the workshop in stimulating the field that might become an occur with increasing frequency the exchange of ideas. Pustejovsky (Brandeis University) presented Different generator designers make that any individual generation his views on the richness of different choices, and the resulting project should define--in its own what can be encoded in what he calls systems are hard to compare.

We recently reported on the application of a machine-learning (ML) technique to automated flight control using a simulated F-16 combat plane (Michie and Camacho 1994). Subsequent tests of our data-induced flying model have broadly confirmed the reported results but have also identified a lack of robustness. We had underestimated the latter and now regard our report (Michie and Camacho 1994) as being, by omission, potentially misleading.

In its early stages, the field of AI had as its main goal the invention of computer programs having the general problem-solving abilities of humans. Along the way, a major shift of emphasis developed from general-purpose programs toward performance programs, ones whose competence was highly specialized and limited to particular areas of expertise. In this article, I claim that AI is now at the beginning of another transition, one that will reinvigorate efforts to build programs of general, humanlike competence. These programs will use specialized performance programs as tools, much like humans do.

Systems (QR '94) was held on 7-10 June A hot issue in cognitive modeling We received 53 submissions and is spatial and diagrammatic reasoning. The core issues of qualitative reasoning Hari Narayanan and his colleagues The eighth workshop was in Nara, included qualitative and (Advanced Research Laboratory, Japan, celebrating the community's causal modeling of the world, automated Hitachi Ltd.) exploited an architecture escape from a simple flip-flop behavior modeling, and qualitative of qualitative visual reasoning and its voyage to a more complex simulation. Interestingly, this transition attracted the attention of many participants. In fact, constructing a component-based sophistication to base qualitative several demonstrations, including model for the input-document handler reasoning on a firm ground. University) presented activity analysis, model abstraction that makes test Iwasaki and Farquhar and will be demonstrating how qualitative generation feasible for continuous held in Monterey, California.

In a companion paper it was shown that the class of constant-depth determinate k-ary recursive clauses is efficiently learnable. In this paper we present negative results showing that any natural generalization of this class is hard to learn in Valiant's model of pac-learnability. In particular, we show that the following program classes are cryptographically hard to learn: programs with an unbounded number of constant-depth linear recursive clauses; programs with one constant-depth determinate clause containing an unbounded number of recursive calls; and programs with one linear recursive clause of constant locality. These results immediately imply the non-learnability of any more general class of programs. We also show that learning a constant-depth determinate program with either two linear recursive clauses or one linear recursive clause and one non-recursive clause is as hard as learning boolean DNF. Together with positive results from the companion paper, these negative results establish a boundary of efficient learnability for recursive function-free clauses.

We study the process of multi-agent reinforcement learning in the context ofload balancing in a distributed system, without use of either centralcoordination or explicit communication. We first define a precise frameworkin which to study adaptive load balancing, important features of which are itsstochastic nature and the purely local information available to individualagents. Given this framework, we show illuminating results on the interplaybetween basic adaptive behavior parameters and their effect on systemefficiency. We then investigate the properties of adaptive load balancing inheterogeneous populations, and address the issue of exploration vs.exploitation in that context. Finally, we show that naive use ofcommunication may not improve, and might even harm system efficiency.