Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

According to W. Edwards Deming, American companies require nothing less than a transformation of management style and of governmental relations with industry. In Out of the Crisis, originally published in 1982, Deming offers a theory of management based on his famous 14 Points for Management. Management's failure to plan for the future, he claims, brings about loss of market, which brings about loss of jobs. Management must be judged not only by the quarterly dividend, but by innovative plans to stay in business, protect investment, ensure future dividends, and provide more jobs through improved product and service. In simple, direct language, he explains the principles of management transformation and how to apply them.

Out of the Crisis

The concept of an agent has become important in both Artificial Intelligence (AI) and mainstream computer science. Our aim in this paper is to point the reader at what we perceive to be the most important theoretical and practical issues associated with the design and construction of intelligent agents. For convenience, we divide these issues into three areas (though as the reader will see, the divisions are at times somewhat arbitrary). Agent theory is concerned with the question of what an agent is, and the use of mathematical formalisms for representing and reasoning about the properties of agents. Agent architectures can be thought of as software engineering models of agents;researchers in this area are primarily concerned with the problem of designing software or hardware systems that will satisfy the properties specified by agent theorists. Finally, agent languages are software systems for programming and experimenting with agents; these languages may embody principles proposed by theorists. The paper is not intended to serve as a tutorial introduction to all the issues mentioned; we hope instead simply to identify the most important issues, and point to work that elaborates on them. The article includes a short review of current and potential applications of agent technology.

/pdf/intelligent-agents-theory-and-practice-5dchj7p8pc.pdf

Intelligent Agents: Theory and Practice

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

From the Publisher: 
The accessible presentation of this book gives both a general view of the entire computer vision enterprise and also offers sufficient detail to be able to build useful applications. Users learn techniques that have proven to be useful by first-hand experience and a wide range of mathematical methods. A CD-ROM with every copy of the text contains source code for programming practice, color images, and illustrative movies. Comprehensive and up-to-date, this book includes essential topics that either reflect practical significance or are of theoretical importance. Topics are discussed in substantial and increasing depth. Application surveys describe numerous important application areas such as image based rendering and digital libraries. Many important algorithms broken down and illustrated in pseudo code. Appropriate for use by engineers as a comprehensive reference to the computer vision enterprise.

Computer vision : a modern approach = 计算机视觉 : 一种现代的方法

We present and analyze a range of techniques for recovering from faults in complex hardware and software systems, from classical techniques that attempt to preserve the abstraction of perfection in the presence of faults to emerging techniques that adapt application functionality to transcend faults, overcome implementation errors in both the hardware and software, and adapt to the characteristics of the underlying execution environment.

/pdf/what-to-do-when-things-go-wrong-recovery-in-complex-computer-4qwkf8jby3.pdf

What to do when things go wrong: recovery in complex (computer) systems

C programmers are all too familiar with out-of-bounds memory errors. A typical scenario involves a pointer calculation that produces an address outside the target block of memory that the developer intends the program to write. The resulting out-of-bounds writes corrupt the data structures of otherwise unrelated parts of the program, causing the program to fail or mysteriously generate unexpected outputs. Analogous errors in Java and other type-safe languages cause out-of-bounds exceptions, which typically terminate the execution of the program. The paper here presents an intriguing technique for automatically isolating and correcting these errors. The basic idea is simple. Allocate blocks randomly within a memory much larger than required to execute the program. Set the remaining unused memory to contain specific \" canary \" values. Run multiple versions of the program to failure, then observe overwritten canary values to compute 1) the target block the out-of-bounds writes were intended to write, and 2) how much additional memory would have been required to bring the writes back within bounds. The fix is to go back to the allocation site in the program that allocated the target block, then apply a patch that suitably increases the size of blocks allocated at that site. The paper also presents a conceptually similar technique for dealing with accesses to prematurely deallocated blocks of memory. Note that the combination of random allocation and multiple heaps is the key insight that makes it possible to isolate the target block. Randomizing the block placement provides the block location diversity required to ensure that, with high probability, only the target block will have the same offset from the overwrit-ten canary values in all of the heaps. One interesting aspect of this approach is that it generates patches almost immediately and without the need for human interaction. It is therefore suitable for uses (such as eliminating security vulnerabilities) that place a premium on obtaining fast responses to newly exposed errors. It can also eliminate the need to interact with a (potentially distracted, indifferent, re-calcitrant, or defunct) software development organization to obtain relief from an error, even when the software is distributed only in binary form. An even more interesting aspect of this approach is that it may very well produce patches that do not completely fix the problem—the fact that the addition of a given amount of memory would have eliminated the out-of-bounds accesses in one observed execution provides no guarantee …

Technical perspectivePatching program errors

Using Dynamic Monitoring to Synthesize Models of Applications That Access Databases

We present a new planning algorithm that formulates the planning problem as a counting satisfiability problem in which the number of available solutions guides the planner deterministically to its goal. In comparison with existing planners, our approach eliminates backtracking and supports efficient incremental planners that add additional sub-formulas without the need to recompute solutions for previously provided subformulas. Our experimental results show that our approach is competitive with existing state-of-the-art planners that formulate the planning problem as a satisfiability problem, then solve the satisfiability problem using specialized off-the-shelf satisfiability solvers such as zChaff

/pdf/incremental-deterministic-planning-1gl9aei54x.pdf

Incremental Deterministic Planning

Runtime verification monitors the execution of a program to determine if it satisfies (typically specified) correctness properties. But what happens when the program violates the correctness properties? The standard view is that continued execution may be unsafe, so the execution must be terminated. We present a variety of intervention techniques that enable software systems to exhibit remarkable flexibility and resilience in the face of errors and faults. These techniques can deliver safe continued execution that offers significant benefits over termination. We also present techniques that build on this malleability to purposefully modify the computation to adapt to changing needs, delivering benefits such as improved performance and reduced power consumption. These results place the advantages of runtime intervention and adaptation clearly on display. They also point the way to a future in which developers produce not the final version of the program which the system blindly executes, but instead a starting point for further modification and evolution as the system adapts to dynamically observed events and conditions.

Martin Rinard

Papers

What to do when things go wrong: recovery in complex (computer) systems

Technical perspectivePatching program errors

Using Dynamic Monitoring to Synthesize Models of Applications That Access Databases

Incremental Deterministic Planning

From Runtime Verification to Runtime Intervention and Adaptation