Present day systems, intelligent or otherwise, are limited by the conceptualizations of the world given to them by their designers. In this paper, we propose a novel, first-principles approach to performing incremental reformulations for computational efficiency. First, we define a reformulation to be a shift in conceptualization: a change in the basic objects, functions, and relations assumed in a formulation. We then analyze the requirements for automating reformulation and show the need for justifying shifts in conceptualization. Inefficient formulations make irrelevant distinctions. A new class of meta-theoretical justifications for a reformulation called irrelevance explanations, is presented. A logical irrelevance explanation demonstrates that certain distinctions made in the formulation are not necessary for the computation of a given class of problems. A computational irrelevance explanation shows that some distinctions are not useful with respect to a given problem solver fo...

Ahn and Lassaline [Ahn, W., Lassaline, M.E., 1995. Causal structure in categorization.

INTRODUCTION Young disciplines often experience moments of doubt: "Are we doing the right thing?" or "Is this approach viable?" [1]. Nowhere is this better exemplified than in the study of computer vision [2]. While progress has been made, the goal of general vision, on the order of human visual perception, remains elusive. Recently, this has led * Please address all correspondence to Michael J. Tarr, P.O. Box 208205, New Haven, CT 06520-8205, E-mail address: tarr@cs.yale.edu to the suggestion that the entire endeavor is flawed, that we should discard the dominant paradigm, and that it should be replaced with a new, more practical alternative. While this position may not qualify as a "paradigm shift" [3], it certainly advocates a substantial change in direction. To justify this radical deviation, proponents of the new, so-called purposive approach muster three lines of support: first, that machines fall far short of the visual capabilities of humans; second, that current com

ositive side, though, the second person might have some advantage over the first person in learning how to shift gears, because the second person would not have to overcome negative transfer from experience with automatic transmissions. As another example, imagine that you are an explorer visiting a remote island, with the purpose of writing a book about the people that you see there. You bring to this island many forms of prior knowledge that will guide you in learning about these new people. For example, based on your experiences in other places, you would expect to see males and females, younger and older people, shy people and arrogant people. You would also have certain hypotheses at a more abstract level, for example, that the clothes that someone wears may be related to the person's age and gender. (Goodman, 1955, referred to such abstract hypotheses as overhypotheses.) In a way, these biases due to previous knowledge might seem to be undesirable. After all, wouldn't be it be be

to all of the category's features. However, people's knowledge of real-world categories often consists of many "rote " features that are not related to their prior knowledge. Five experiments found that even minimal prior knowledge (1 knowledge-relevant feature and 5 rote features per exemplar) can facilitate category learning. Posttests revealed that although the knowledge aided learning, subjects also acquired the rote features that were not related to knowledge, contradicting predictions of an attentional expla-nation of the knowledge effect. The results of Experiment 6 suggested that subjects attempt to link even rote features to their knowledge.

Theories of categorization have typically focused on the internal structure of categories. This paper is concerned with the external structure of categories. In particular , it is suggested that many categories specify the relational role that is played by category members. To support this claim, the paper distinguishes between traditional feature-based categories, relational categories (which specify a relational structure) and role-governed categories (which specify that an item plays a particular role within a relational structure). After discussing the relationship among these types of categories, the implications of this view for the study of category learning and category use are discussed.

this paper is to show that the category inclusion principle has only limited descriptive validity

1 Comprehensive visual learning is the treatment of theories and techniques for computer vision systems to automatically learn to understand comprehensive visual information with minimal human-imposed rules about the visual world. This article discusses some major performance difficulties encountered by currently prevailing approaches to computer vision and introduces the promising direction of comprehensive learning towards overcoming these difficulties. It also indicates why the direction may have a profound impact on the performance of computer vision algorithms for real world problems. Some example techniques for comprehensive visual learning are presented. 1 Introduction An image of a real-world scene depends on a series of factors, illumination, object shape, surface reflectance, viewing geometry, sensor type, etc. The image is a result of compound interactions among these factors. In the real world, change in these factors is ubiquitous and mostly is not known a priori. This ...

Human concept learning presents a version of the classic problem of induction, which is made particularly difficult by the combination of two requirements: the need to learn from a rich (i.e. nested and overlapping) vocabulary of possible concepts and the need to be able to generalize concepts reasonably from only a few positive examples. I begin this thesis by considering a simple number concept game as a concrete illustration of this ability. On this task, human learners can with reasonable confidence lock in on one out of a billion billion billion logically possible concepts, after seeing only four positive examples of the concept, and can generalize informatively after seeing just a single example. Neither of the two classic approaches to inductive inference -- hypothesis testing in a constrained space of possible rules and computing similarity to the observed examples -- can provide a complete picture of how people generalize concepts in even this simple setting. This thesis prop...

Experiments investigated whether infants infer that a hidden, freely moving object will move continuously and smoothly. Infants aged 6 and 10 months, like the $-month-old infants in previous experiments, inferred that the object’s path would be connected and unobstructed, in accord with the principle of continuity. In contrast, 4- and 6-month-old infants did not appear to infer that the object’s path would be smooth, in accord with the principle of inertia. At 8 and 10 months, knowledge of inertia appeared to be emerging but remained weaker than knowledge of continuity. These findings are consistent with the view that common sense knowledge of physical objects develops by enrichment around constant core principles. The core knowledge thesis Human adults generally can predict how the things around them will behave. When a ball rolls from view on a table, for example, adults infer that it will continue to exist and to move on a connected path, that it will move smoothly in the absence of obstacles or surface irregularities, that it will rebound from or Supported by grants from NSF (BNS-8613390) and NIH (HD-23103) and by a fellowship to E.S.S. from the John Simon Guggenheim Memorial Foundation. We thank Frank Keil for comments and