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

物件導向軟體之架構(Object-Oriented Software Construction)探討

This paper studies the control of a class of discrete event processes, i.e. processes that are discrete, asynchronous and possibly nondeter-ministic. The controlled process is described as the generator of a formal language, while the controller, or supervisor, is constructed from the grammar of a specified target language that incorporates the desired closed-loop system behavior. The existence problem for a supervisor is reduced to finding the largest controllable language contained in a given legal language. Two examples are provided.

/pdf/supervisory-control-of-a-class-of-discrete-event-processes-45fsmhdhbo.pdf

Supervisory control of a class of discrete event processes

Since the subject of traffic dynamics has captured the interest of physicists, many surprising effects have been revealed and explained. Some of the questions now understood are the following: Why are vehicles sometimes stopped by ``phantom traffic jams'' even though drivers all like to drive fast? What are the mechanisms behind stop-and-go traffic? Why are there several different kinds of congestion, and how are they related? Why do most traffic jams occur considerably before the road capacity is reached? Can a temporary reduction in the volume of traffic cause a lasting traffic jam? Under which conditions can speed limits speed up traffic? Why do pedestrians moving in opposite directions normally organize into lanes, while similar systems ``freeze by heating''? All of these questions have been answered by applying and extending methods from statistical physics and nonlinear dynamics to self-driven many-particle systems. This article considers the empirical data and then reviews the main approaches to modeling pedestrian and vehicle traffic. These include microscopic (particle-based), mesoscopic (gas-kinetic), and macroscopic (fluid-dynamic) models. Attention is also paid to the formulation of a micro-macro link, to aspects of universality, and to other unifying concepts, such as a general modeling framework for self-driven many-particle systems, including spin systems. While the primary focus is upon vehicle and pedestrian traffic, applications to biological or socio-economic systems such as bacterial colonies, flocks of birds, panics, and stock market dynamics are touched upon as well.

Traffic and related self-driven many-particle systems

https://www.ufz.de/export/data/2/100065_GrimmODD.pdf

A standard protocol for describing individual-based and agent-based models

An 'evaluation' approach devised for an inductive reasoning system called logic-based discrete-event inductive reasoner is the focus of this paper. The underlying inductive reasoning methodology utilizes abstractions as its primary means to deal with lack of knowledge. Based on abstractions and their treatments as assumptions, the logic-based discrete-event inductive reasoning system allows non- monotonic predictions. The evaluation approach takes into account explicitly the role of abstractions employed in non- monotonically derived multiple predictions. These predictions are ranked according to the type and number of abstractions used. The proposed evaluation approach is also discussed in relation to the dichotomy of model validation and simulation correctness.© (1998) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

/pdf/evaluating-model-abstractions-a-quantitative-approach-2vi3c1nuyd.pdf

Evaluating model abstractions : A quantitative approach

We describe how to develop a suite of models in the MS4 Modeling Environment. The approach employs the operation of merging of System Entity Structures supported by the environment. After construction, the suite of models can be hosted on Model Store, the cloud-based repository of models provided by MS4 systems as a basis for further collaborative model development. A suite of models, relating to Global Warming is used as an example. We discuss enhanced browsing as necessary to enable a developer to work effectively in a marketplace environment.

Creating suites of models with system entity structure: global warming example

Computer-based methodologies for systems modelling, design, and analysis require the user to understand and manipulate abstractions of the objects of direct interest. The success of such methodologies may hang on human ability to deal with such abstractions and the degree to which the computer assistance meshes with this ability. This paper formulates systems methodologies within a problem solving frame-work in which abstractions, and the role they play, are formally represented. Two kinds of abstraction-based methodologies are distinguished: information-based methodologies employ abstractions to provide projections of the current state of the problem solving process, subsearch methodologies augment the underlying solution space with abstract representations that are intended to facilitate better perception of goal directions and improved quality of solutions. In this framework, implementation of abstraction classes is shown to involve overhead costs to both software and “brainware” that may vitiate the intended benefits. Components of, and approaches to, such a cost/benefit analysis are presented.

Abstraction in methodology: A framework for computer support

The use of activity-based credit assignment (ACA) for the automatic evaluation and selection of candidate components of systems is considered here. The whole process consists of a precise automatic structured specification of systems. Mathematical definitions and algorithms are provided. ACA converges on good components/compositions faster than repository-based random search. As systems constitute a vast class of problems to be specified by a modeler, this automatic composition of systems opens new research perspectives. The paper also places ACA within the context of existing approaches to credit assignment in classifier systems.

https://hal.archives-ouvertes.fr/hal-00674306/document

Bernard P. Zeigler

Papers

Evaluating model abstractions : A quantitative approach

Creating suites of models with system entity structure: global warming example

Abstraction in methodology: A framework for computer support

Activity-based credit assignment (ACA) in hierarchical simulation

AI, Simulation and Planning in High Autonomy Systems (Conference, '91)