Traditional AI Research Articles

Angiogenic Index: A New Method for Assessing Microvascularity in Breast Carcinoma with Possible Prognostic Implications.

The quantitation of microvessels in breast cancer in sections immunolabeled for factor VIII, CD-34, or CD-31 currently is expressed as numbers of vessels/mm2. In an attempt to perform a more accurate method for counting microvessels, we determined the number of microvessels by 1,000 tumor cells (angiogenic index; AI). This new method minimizes the possible variations concerning the width of the microscopic fields, stroma/epithelium relations, and cellular tumor size. The present study compares the number of vessels determined by the traditional method (vessels/mm2) and AI in a series of 215 cases of ductal infiltrating carcinoma not otherwise specified. Also we studied the degree of correlation between both methods and with pathologic variables (tumor size, histologic grade, mitotic count, tumor necrosis, vascular invasion, skin infiltration, and axillary lymph node metastasis). Our results showed that the AI correlates more significantly than the classic microvessel density determination with other prognostic factors in breast cancer. In histologic grade III tumors, high AI correlates significantly with the presence of more than three axillary lymph nodes metastases. Therefore we recommend determining the degree of tumor angiogenesis by counting vessels per 1,000 tumor cells (AI) because of its reliable determination and significant correlation with other prognostic factors.

Utility Models for Goal‐Directed, Decision‐Theoretic Planners

AI planning agents are goal‐directed: success is measured in terms of whether an input goal is satisfied. The goal gives structure to the planning problem, and planning representations and algorithms have been designed to exploit that structure. Strict goal satisfaction may be an unacceptably restrictive measure of good behavior, however.A general decision‐theoretic agent, on the other hand, has no explicit goals: success is measured in terms of an arbitrary preference model or utility function defined over plan outcomes. Although it is a very general and powerful model of problem solving, decision‐theoretic choice lacks structure, which can make it difficult to develop effective plan‐generation algorithms.This paper establishes a middle ground between the two models. We extend the traditional AI goal model in several directions: allowing goals with temporal extent, expressing preferences over partial satisfaction of goals, and balancing goal satisfaction against the cost of the resources consumed in service of the goals. In doing so we provide a utility model for a goal‐directed agent.An important quality of the proposed model is its tractability. We claim that our model, like classical goal models, makes problem structure explicit. This structure can then be exploited by a problem‐solving algorithm. We support this claim by reporting on two implemented planning systems that adopt and exploit our model.

Using mechanical hardware to simplify process planning

Process planning is an important and complex task in manufacturing. In this paper we present the view that process planning is difficult to automate because of the state of machining technology today. Many standard practices in machining, like fixturing, are really specific solutions that have evolved over the years for particular problems; modeling this patchwork of technologies, and planning with them is inherently difficult. To show this, we attempt to cast process planning for machining in the form of traditional AI planning problem using predicate calculus. Specifically, we identify a class of interactions, known as global interactions, that make the resulting planning task very difficult. We show how many of these global interactions can be approximated as pairwise ordering constraints that are amenable to well known nonlinear planning techniques. However, some global interactions, particularly those arising from fixturing and rigidity concerns, do not lend themselves to such simplification. To handle these interactions we propose that a paradigm-shift is required in machining technology. We describe a new fixturing technology called reference free part encapsulation (RFPE) that makes it posssible to manufacture almost arbitrarily complex components. RFPE makes it possible to ignore global interactions and therefore justifies the reduction of general process planning to a nonlinear planning problem.

Coordinating the eyes, head and arm of an autonomous robot

This paper presents an approach to guide a robot arm visually, by using an active stereo camera mounted on an active head platform. The camera and head system is capable of changing the look-direction and fixating on the target object. Unlike the traditional AI approach, the proposed approach does not require full perspective camera calibration or scene reconstruction. Visual parameters which contain relationships with the robot or world space are used instead of the reconstructed three-dimensional (3D) world information. The necessary visual cues required to control the robot are developed. These visual cues are computed solely from the image coordinates. The relationship between these visual cues and the robot space is then established via a mapping matrix. Using this relation together with visual feedback, the end-effector can be guided towards the target object. Even though the 3D world coordinates are not recovered, the positioning accuracy attainable is high. To allow for vergence movement of the stereo camera or pan–tilt movements of the head platform, methods to compensate the mapping matrix are proposed. This approach eliminates the need for re-calibration after any deliberate change in the configuration of the head platform. Therefore, building an autonomous robot with coordinated head, eye and hand becomes technically feasible. Simulation studies on the capability of the system to reach arbitrary target objects are presented. High flexibility and large accessible workspace are the main advantages of the proposed autonomous robot system.

The robocup physical agent challenge: Phase i

Traditional AI research has not given due attention to the important role that physical bodies play for agents as their interactions produce complex emergent behaviors to achieve goals in the dynamic real world. The RoboCup Physical Agent Challenge provides a good test bed for studying how physical bodies play a significant role in realizing intelligent behaviors using the RoboCup framework (Kitano et al., 1995). In order for the robots to play a soccer game reasonably well, a wide range of technologies needs to be integrated, and a number of technical breakthroughs must be made. In this article, we present three challenging tasks as the RoboCup Physical Agent Challenge Phase I: (1) moving the ball to the specified area (shooting, passing, and dribbling) with no, with stationary, or with moving obstacles; (2) catching the ball from an opponent or a teammate (receiving, goal keeping, and intercepting); and (3) passing the ball between two players. The first two tasks are concerned with single-agent skills, while the third is related to a simple cooperative behavior. Motivation for these challenges and evaluation methodology is given.

Applying plan recognition algorithms to program understanding

Program understanding is often viewed as the task of extracting plans and design goals from program source. As such, it is natural to try to apply standard AI plan recognition techniques to the program understanding problem. Yet program understanding researchers have quietly, but consistently, avoided the use of these plan recognition algorithms. This paper shows that treating program understanding as plan recognition is too simplistic and that traditional AI search algorithms for plan recognition are not suitable. In particular, we show that: the program understanding task differs significantly from the typical general plan recognition task along several key dimensions; the program understanding task has particular properties that make it particularly amenable to constraint satisfaction techniques; and augmenting AI plan recognition algorithms with these techniques can lead to effective solutions for the program understanding problem.

Concurrent competitive wave approach to hyper-distributed hyper-parallel AI processing

This paper presents a new approach of the synchronous homogeneous concurrent propagation of competitive waves for the purpose of hyper-distributed hyper-parallel heuristic problem-solving. The concurrent algorithm, mechanism and their properties are given. In comparison with the traditional AI algorithms, the approach is featured by the knowledge-based problem-solving in the distributed parallel environment, the feasibility for hardware implementation and the various applications.

Vision based modeling for film and multimedia production

Recovery of a three-dimensional (3-D) description of the world from images has been a central topic in computer vision from the early days of AI research and is currently becoming of great importance, particularly to the film and entertainment industries. In particular, this article addresses the problems of preproduction and postproduction for 3-D compositing of graphics in film and of building multimedia content for computer entertainment systems, both areas of burgeoning activity currently driven extensively with expensive skilled human labor. In this article, we discuss how the particular requirements for accuracy and generality of these applications impact the choices of vision techniques and how the interactive nature of production systems can be exploited to bypass the difficulties of segmentation and scene understanding that has prevented equally successful systems from arising in traditional AI contexts. We show actual results of both preproduction and postproduction processes in film and multimedia production, including preproduction model building, postproduction alignment of 3-D mattes into film, and a combination of preproduction and postproduction for inserting 3-D graphics into a video scene. The article concludes with a discussion of how the technology can be profitably merged with existing production practices and how it can be extended to provide further capabilities in production of entertainment media.

The subtlety of sameness: a theory and computer model of analogy-making

The research described in this book is based on the premise that human analogy-making is an extension of our constant background process of perceiving - in other words, that analogy-making and the perception of sameness are two sides of the same coin. At the heart of the author's theory and computer model of analogy-making is the idea that the building-up and the manipulation of representations are inseparable aspects of mental functioning, in contrast to traditional AI models of high-level cognitive processes, which have almost always depended on a clean separation. A computer program called Tabletop forms analogies in a microdomain consisting of everyday objects on a table set for a meal. The theory and the program rely on the idea that myriad stochastic choices made on the microlevel can add up to statistical robustness on a macrolevel. To illustrate this, French includes the results of thousands of runs of his program on several dozen interrelated analogy problems in the Tabletop microworld.

Speeding up problem solving by abstraction: a graph oriented approach

This paper presents a new perspective on the traditional AI task of problem solving and the techniques of abstraction and refinement. The new perspective is based on the well-known, but little exploited, relation between problem solving and the task of finding a path in a graph between two given nodes. The graph oriented view of abstraction suggests two new families of abstraction techniques, algebraic abstraction and STAR abstraction. The first is shown to be extremely sensitive to the exact manner in which problems are represented. STAR abstraction, by contrast, is very widely applicable and leads to significant speedup in all our experiments. The reformulation of traditional refinement techniques as graph algorithms suggests several enhancements, including an optimal refinement algorithm, and one radically new technique: alternating search direction. Experiments comparing these techniques on a variety of problems show that alternating opportunism (AltO) a variant of the new technique, is uniformly superior to all the others.

Agents with emotions in behavioral animation

This paper presents an innovative architecture for emotional characters in computer animation based on a reactive agent structure. The architecture is based on a cognitive model where both controlled and automatic procedures coexist. The proposed architecture breaks with traditional AI paradigms and establishes an efficient approach to develop behavioral animation systems.

A Distributed Approach to Block Stacking Problem Based on Evolutionary Learning.

We describe a new approach to a nonlinear block stacking problem in a traditional AI field. In our problem setting, we see the problem as one of planning in a dynamic environment and attempt to solve it in a distributed manner. Each block is considered as an autonomous agent having sensors and effectors and capable of moving in the block world. As the behaviors of the agents change the environment, the agents are required to resolve conflict among other agents. In our approach, motions of the agents are controlled using a recurrent neural network (RNN). Based on sensory inputs to the RNN, the agents select their alternative actions. The RNN implemented in each agent is trained through evolutionary learning and both the structure and connection weights of the network are altered appropriately using two kinds of mutations according to the degree of success of the planning. Actions of agents are triggered by activation levels which are calculated on the basis of received signals from other agents and the environment. Based on the proposed methodology, computational experiments are carried out and some experimental results are shown.

The ?explicit-implicit? distinction

Much of traditional AI exemplifies the “explicit representation” paradigm, and during the late 1980's a heated debate arose between the classical and connectionist camps as to whether beliefs and rules receive an explicit or implicit representation in human cognition. In a recent paper, Kirsh (1990) questions the coherence of the fundamental distinction underlying this debate. He argues that our basic intuitions concerning ‘explicit’ and ‘implicit’ representations are not only confused but inconsistent. Ultimately, Kirsh proposes a new formulation of the distinction, based upon the criterion ofconstant time processing.

A formal semantics for DAI language NUML

Traditional AI systems are brittle in the sense that they fail miserably when presented with problems even slightly outside of their limited range of expertise. A powerful, extensible strategy of Distributed Artificial Intelligence (DAI) for overcoming such bounds is to put the system in a society of systems. So the ability to coordinate group activities of individuals and to communicate between each other is necessary for a language describing DAI systems. Agent-oriented language NUML is such a language. It is a specific kind of object-oriented language. To give formal semantics to NUML, there is the problem to formalise object-oriented programming paradigm which is still open. The theory of higher-order π-calculus is a concurrent computation model with sufficient capability, which provides us a mathematical tool to do the formalization. This paper tries to use higher-order π-calculus to formalise NUML.

On Babies and Bathwater: A Cautionary Tale

■ One should not throw out the baby with the bathwater, according to an old aphorism. Some popular recent positions in AI thinking have done just this, we suggest, by rejecting the useful idea of mental representations in their overenthusiastic zeal to correct some simplifications and naiveties in the way traditional AI ideas have sometimes been understood. These “situated” perspectives correctly emphasize that agents live in a social world, using their environments to help guide their actions without needing to always plan their futures in detail; but they incorrectly conclude that the very idea of mental representation is mistaken. This perspective has its intellectual roots in parts of recent sociological thinking which reject the entire fabric of western science. We discuss these ideas and disputes in the form of an illustrated fable concerning nannies and babies.

A connectionist approach to diagnostic problem solving using causal networks

In general diagnostic problem solving using causal networks, one is often interested in identifying the most plausible explanation for observed manifestations. Although a number of sequential AI search methods have been applied to this task, these methods suffer from potential combinatorial explosion. It thus seems reasonable to explore whether connectionist modeling methods can overcome this difficulty through the use of extensive explicit parallel processing. A number of studies have recently implemented diagnostic systems using connectionist models, but these are based on a distributed representation and face significant limitations that are summarized in this paper. Accordingly, we have developed a connectionist model of diagnostic inference using a local representation to address these limitations. This model controls spreading activation in a causal network that has rapidly varying connection strengths. It generates diagnostic hypotheses based on the same causal network that would be used by a traditional AI search algorithm. Experimental results are presented that demonstrate that this connectionist model can closely approximate the best diagnostic hypotheses with some causal networks, even when multiple disorders are present simultaneously.

A connectionist and a traditional AI quantizer, symbolic versus sub-symbolic models of rhythm perception

The Symbolic AI paradigm and the Connectionist paradigm have produced some incompatible models of the same domain of cognition. Two such models in the field of rhythm perception, namely the Longuet-Higgins Musical Parser and the Desain & Honing connectionist quantizer, were studied in order to find ways to compare and evaluate them. Different perspectives from which to describe their behavior were developed, providing a conceptual as well as a visual representation of the operation of the models. With these tools it proved possible to discuss their similarities and differences and to narrow the gap between sub-symbolic and symbolic models.

An application-based agenda for incorporating OR into an AI design environment for facility design

As part of our study of expert human problem solving in design, we devised and implemented the core of a workstation-based, graphics-intensive, intellegent support system for facility planning and layout. Our experience, in that endeavor, is that neither traditional AI nor current OR/MS approaches are adequate to provide a complete design support system, but that a hybrid system might be. In this paper, we describe how such a hybridation could occur. We identify points in the problem-solving process at which appropriate AI mechanisms can be brought to bear to assist in the selection, formulation, and interpretation of existing numeric (OR/MS) methods. We conclude that it will be necessary to relax traditionalm OR/MS assumptions about the nature of the facility planning problem leg., that the problem is sequential, with facility specification followed by facility layout) and about the proper relationship between the designer/planner and the design tool. The non-sequential nature of facility design suggests that (a) a design support system will need to work in an interactive, as opposed to a batch, mode, (b) the design support system must be capable of formulating the problem to be solved as part of the solution process, as is done by the human expert, and (c) the burden of solving the problem should not be shifted to the system. Building and testing the hybrid system remains to be done.

The second international workshop on human and machine cognition

The Second International Workshop on Human and Machine Cognition was held on 9-11 May 1991. Participation was limited to 40 researchers who are principally involved in computer science, philosophy, and psychology. The workshop focused on the foundational and methodological concerns of those who want to forge a robust and scientifically respectable AI and cognitive science. With the theme of What do androids know, and when do they know it? the positions covered a wide range and presented a lot of room for disagreement. The debate between the traditional AI and the situated cognition types and the connnectionists was a focal point for discussion during the workshop.

Prior knowledge strengthens learning to control search in weak theory domains

Rather than search exhaustively in large problem spaces, people use heuristics that entail minimal search and yet support intelligent decisions. Human performance in such spaces is robust and improves with experience. Game playing is a good example. the traditional AI approach to game playing, however, has been to build a program that plays only a single game, and plays it very well. Because exhaustive search of the game graph and minimax of the resultant values guarantees perfect play, programmed champions rely on deep, fast, efficient search. Despite their more limited, fallible memories and arguably slower processors, human masters often defeat a program whose search is incomplete. People search less, remember less, and somehow make the right choices. This article describes HOYLE, a program that learns to play specific games under the direction of a weak theory, clearly delineated prior knowledge about games in general. HOYLE differs from most game-playing programs in two ways: it is able to play any of a broad class of games according to the rules, and it improves its performance through a variety of learning paradigms. HOYLE shows that, for a class of related problems, it is possible to replace deep search with a weak theory based upon heuristically selective cache memories and a consensus about action among reliable rationales.