Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

The advent of software agents gave rise to much discussion of just what such an agent is, and of how they differ from programs in general. Here we propose a formal definition of an autonomous agent which clearly distinguishes a software agent from just any program. We also offer the beginnings of a natural kinds taxonomy of autonomous agents, and discuss possibilities for further classification. Finally, we discuss subagents and multiagent systems.

/pdf/is-it-an-agent-or-just-a-program-a-taxonomy-for-autonomous-oncwlzq0wr.pdf

Is it an Agent, or Just a Program?: A Taxonomy for Autonomous Agents

Let's read! We will often find out this sentence everywhere. When still being a kid, mom used to order us to always read, so did the teacher. Some books are fully read in a week and we need the obligation to support reading. What about now? Do you still love reading? Is reading only for you who have obligation? Absolutely not! We here offer you a new book enPDFd the perception of the visual world to read.

The Perception of the Visual World. By James J. Gibson. U.S.A.: Houghton Mifflin Company, 1950 (George Allen & Unwin, Ltd., London). Price 35s.

Systems and methods for navigating hypermedia using multiple coordinated input/output device sets. Disclosed systems and methods allow a user and/or an author to control what resources are presented on which device sets (whether they are integrated or not), and provide for coordinating browsing activities to enable such a user interface to be employed across multiple independent systems. Disclosed systems and methods also support new and enriched aspects and applications of hypermedia browsing and related business activities.

Method and apparatus for browsing using multiple coordinated device sets

Method and Apparatus for Browsing Using Alternative Linkbases

This paper describes design criteria for creating highly embedded, interactive spaces that we call Intelligent Environments. The motivation for building these systems is to bring computation into the real, physical world to support what is traditionally considered non-computational activity. We describe an existing prototype space, known as the Intelligent Room, which was created to experiment with different forms of natural, multimodal human-computer interaction. We discuss design decisions encountered while creating the Intelligent Room and how the experiences gained during its use have shaped the creation of its successor.

/pdf/design-principles-for-intelligent-environments-mwhlcok0e9.pdf

Design principles for intelligent environments

Intelligent Environments (IEs) have specific computational properties that generally distinguish them from other computational systems. They have large numbers of hardware and software components that need to be interconnected. Their infrastructures tend to be highly distributed, reflecting both the distributed nature of the real world and the IEs’ need for large amounts of computational power. They also tend to be highly dynamic and require reconfiguration and resource management on the fly as their components and inhabitants change, and as they adjust their operation to suit the learned preferences of their users. Because IEs generally have multimodal interfaces, they also usually have high degrees of parallelism for resolving multiple, simultaneous events. Finally, debugging IEs present unique challenges to their creators, not only because of their distributed parallelism, but also because of the difficulty of pinning down their “state” in a formal computational sense. This paper describes Metaglue, an extension to the Java programming language for building software agent systems for controlling Intelligent Environments that has been specifically designed to address these needs. Metaglue has been developed as part of the MIT Artificial Intelligence Lab’s Intelligent Room Project, which has spent the past four years designing Intelligent Environments for research in Human-Computer Interaction.

/pdf/meeting-the-computational-needs-of-intelligent-environments-1x6c3buh53.pdf

Meeting the Computational Needs of Intelligent Environments: The Metaglue System

This paper argues that complex, embedded software agent systems are best constructed with parallel, layered architectures. These systems resemble Minskian Societies of Mind and Brooksian subsumption controllers for robots, and they demonstrate that complex behaviors can be had via the aggregates of relatively simple interacting agents. We illustrate this principle with a distributed software agent system that controls the behavior of our laboratory's Intelligent Room.

/pdf/building-brains-for-rooms-designing-distributed-software-1knggdfvxb.pdf

Building brains for rooms: designing distributed software agents

This thesis presents {\em SodaBot}, a general-purpose software agent user- environment and construction system. Its primary component is the {\em basic software agent} --- a computational framework for building agents which is essentially an {\em agent operating system}. We also present a new language for programming the basic software agent whose primitives are designed around human-level descriptions of agent activity. Via this programming language, {\em users can easily implement a wide-range of typical software agent applications}, e.g. personal on-line assistants and meeting scheduling agents. The SodaBot system has been implemented and tested, and its description comprises the bulk of this thesis.

SodaBot: a software agent environment and construction system

he creation of software agents has generated much recent interest. A variety of approaches and prw jects use the term \" agents \"-ranging from adaptive user interfaces (W) to systems that use planning algorithms to generate shell scripts [ 1, 2, 41. In our own approach, agents assist users in a range of daily, mundane activities-setting up meetings, sending out papers, locating information in multiple databases, tracking the whereabouts of people, and so on. Our objective is to design agents that blend transparently into normal work environments, while relieving users of low-level adminis trative and clerical tasks. The practical aspects of software agents are taken seriously. Users should be able to assume the agents are reliable and predictable and the human user remains in ultimate control. One of the most difficult aspects of agent design is to define specific tasks that are both feasible using current technology, and are truly useful to the everyday user. Furthermore, it became clear during the testing of our initial protm type that users have little patience when it comes to interacting with software agents. We therefore paid special attention to the Ul aspects of our system. In particular, whenever possible, we opted for graphically oriented interfaces over pure text-based interfaces. In addition, reliability and error handling is crucial in all parts of a software agent system. The real world is unpredictable: messages between agents may be lost or delayed, pea-pie may respond inappropriately rounding the scheduling of a visitor to our laboratory. We designed and implemented a set of software agents to handle this task. We deliberately made no commitment in advance to a particular agent architecture. We then tested the system with ordinary users, providing feedback that led to many improvements in the design of the agents and the human/ agent interfaces, as well as to the development of a general and flexible framework for agent interaction. The key feature of the framework is the use of personal-ized agents called \" userbots \" that mediate communication between users and task-specific agents. We believe the bottom-up approach is crucial in identifying the necessary properties of a successful agent platform. Our initial experiments have already led us to formulate some key properties. Examples include the separation of task-specific agents from user-centered agents, the need to handle issues of security and privacy, and as mentioned earlier, the need for good human interfaces and high reliability. TaSkbOts …

/pdf/bottom-up-design-of-software-agents-4kqjfdstlu.pdf

Michael H. Coen

Papers

Design principles for intelligent environments

Meeting the Computational Needs of Intelligent Environments: The Metaglue System

Building brains for rooms: designing distributed software agents

SodaBot: a software agent environment and construction system

Bottom-up design of software agents