Michalski, R.S. (Ed.) (1993): Multistrategy Learning. Kluwer Academic Publishers.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....that addresses the CMOMMT problem without the assumption of an a priori model. This approach combines reinforcement learning, lazy learning, and a pessimistic algorithm able to compute for each team member a lower bound on the utility of executing an action in a given situation. Lazy learning [1] is used to enable robot team members to build a memory of situation action pairs through random exploration of the CMOMMT problem. A reinforcement function gives the utility of a given situation, providing positive feedback if new targets are acquired and negative feedback if targets are lost. ....

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

....most used mechanisms consists of splitting other par titions, and reassigning the new partition to the empty one. All empty cells generated by the GLA are changed in each iteration by another cell. To define he new one, another non empty cell, y, wih big average dis orion is spli in wo: Yl: y[1] , y[K] e ,and Initial codebook generation. We have used a version of the GLA [15] as explained in Table 2, that requires a partition split mechanism as the one described above inserted into the GLA in Table 1. 2.2 Q learning Q learning [25, 26] is one of the most used reinforcement ....

....to nearby targets and is repulsed by nearby robots, with the movement of the robot calculated as the weighted summation of attractive and repulsive force vectors. The learning approach in this previous research, called Pessimistic Lazy Q Learning, is based on a combination of lazy learning [1], Q Learning, and a pessimistic algorithm for evaluating global rewards. This instance based algo rithm stores a set of situations in a memory in order to use them when needed. A pessimistic utility metric is used to choose the right action from this set of situations. Fig. 3 compares the results ....

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

....DCT algorithms with respect to increasingly important design metrics such as power con sumption and latency. 2. 0 Evaluation Strategy and Experimental Set Up For synthesis and algorithm space exploration we used the high level synthesis system, FlYPER lRab91] and the silicon compiler, LAGER [Bro92] from the University of California at Berkeley. We selected the HYPER high level synthesis tool because it provides exceptionally strong support for design space exploration. We coded all DCT algorithms in the applicative, functional, DSP programming language SILAGE which HYPER translates into the ....

R. W. Brodersen, editor. Anatomy of a Silicon Compiler. Kluwer Academic Publishers, 1992.

....D1 D1 C 1 C 2 C n Performance Estimator Feature subset and classifier Object Definition Hierarchy in the set j. In binary classifiers, c j :U# 0,1 where, # i#U , c j (i) 1 if i#j and c j (i) 0 if i#j. In fuzzy classifiers 35 the function is not binary, but continuous, thus c :U#[0,1]. In this case, j is a fuzzy set since each element in j has been assigned a degree of membership in that set (e.g. if c j (i) 0.75 and c j (l) 0.68 we say that i is a stronger member of class j than l) Region classification results in a set of region membership tuples R op = r 1 , m 1 , s ....

D. Aha, Editor. Lazy Learning. Kluwer Academic Publishers, The Netherlands, 1997.

....logic; cf. e.g. Jones and Sergot, 1992 ] From our perspective this issue is not very relevant, since logics for defeasible argumentation can cope with any underlying logic. Moreover, as for the defeasibility of deontic reasoning, we think that special deontic defeasible logics (see e.g. Nute, 1997 ] are not very suited. It is better to embed one s preferred deontic monotonic logic in one s preferred general defeasible logic, since legal defeasibility is not restricted to deontic terms, but extends to all other kinds of legal knowledge, including de nitions and evidential knowledge. ....

D. Nute, editor. Defeasible Deontic Logic, volume 263 of Synthese Library. Kluwer Academic Publishers, Dordrecht/Boston/London, 1997.

....In particular, it should be suited for the medical experts to build large networks. An other point is that we expect that the final network will be too complex for exact reasoning. Therefore the system should be able to do approximate reasoning, e.g. using mean field or variational methods[1, 2, 3]. To optimally meet (future) requirements of the users, and in anticipation to the need of (future) approximating methods, we constructed our own software system, called BayesBuilder. Currently, BayesBuilder has most of the standard features, such as the junction tree algorithm; the possibility ....

M. Jordan, editor. Learning in Graphical Models. Kluwer Academic Publishers, 1996. NATO ASI Series.

....databases [AHV95] knowledge representation [Bre96] arti cial intelligence [GHR93] formal languages [RS97] etc. Going in the opposite direction, from computer science to logic, we nd extremely fast implementations of model checkers and tableaux based and resolution based theorem provers [BS98], automatatheoretic methods for deciding powerful languages [BK98] the tight connections between the theories of computational and descriptive complexity [Imm99] etc. And this is just a small part of a far bigger development, as logic continues to play an important role in computer science and ....

....and application driven results, which is almost a hallmark of the literature on description logics. Resolution Based Methods for Modal Logics by Hans de Nivelle, Renate Schmidt, and Ullrich Hustadt. Automated theorem proving for rst order logics is a well developed eld with years of history [BS98]. Attending a conference on the eld, like for example CADE, Computer Aided Deduction, can be an unforgettable experience, where highly tuned heuristics will be discussed and test beds will be crunched down by provers trying to outperform each other. Now, the standard or relational ranslation ....

W. Bibel and P.H. Schmitt, editors. Automated Deduction { A Basis for Applications. Kluwer Academic Publishers, 1998.

....an action in a given situation. The challenges in this multi robot learning problem include a very large search space, the need for communication or awareness of robot team members, and the diculty of assigning credit in an inherently cooperative problem. In this learning approach, lazy learning [20] is used to enable robot team members to build a memory of situation action pairs through random exploration of the CMOMMT problem. A reinforcement function gives the utility of a given situation. The pessimistic algorithm for each robot then uses the utility values to select the action that ....

....Markov Decision Process (POMDP) domain) the existence of local indicators that help individual robots perform their tasks, nor the use of symbolic representations. The following subsections describe this approach and its results in more detail. Lazy learning and Q learning Lazy learning [20] also called instance based learning promotes the principle of delaying the use of the gathered information until the necessity arises (see Fig. 4) The same pool of information (i.e. memory) is used for di erent behavior syntheses. The lazy memory provides a good way of reducing the duration ....

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

....of manufacturing system. These methods have a significant impact when agents have to make a choice among a number of machines. Four functions are investigated in this paper: 1) Expected value function 2 )Expected utility function 3) Rank dependent expected utility function 4) Prospect function [5, 6, 13]. In the previous research, the authors used price and shape factor curves to compute the utility values for part and machine agents for making decisions [7, 8] This research replaces those curves with a dynamic utility function which depends on current buffer size, estimated processing time ....

W. Edwards, editor. Utility Theories: Measurements and Applications. Studies in Risk and Uncertainty. Kluwer Academic Publishers, 1992.

....Srtv IS T SPIE Storage and Retrieval for Image and Video Databases VIII, Vol. 3972, San Jose, CA, Jan. 2000 of other labels) The training set was used to build classifiers and 10 fold cross validation was done to determine the their performance. IB 1 IB 3 IB 5 ID3 Nave Bayes MC4 Pitcher [0, 1, 2, 4, 11, 12] CV: 0.43 T: 22.75 R: 116 [1, 2, 4, 7, 12, 15, 17, 37] CV: 0.43 T: 23.63 R: 118 [1, 2, 7, 12, 20, 37] CV: 0.58 T: 23.14 R: 115 1, 2, 7, 9, 11, 12, 19, 20, 37] CV: 1.30 T: 22.36 R: 115 [1, 2, 3, 6, 12, 18, 21, 30, 37] CV: 1.29 T: 17.60 R: 112 [1, 2, 10, 11, 15, ....

....and Video Databases VIII, Vol. 3972, San Jose, CA, Jan. 2000 of other labels) The training set was used to build classifiers and 10 fold cross validation was done to determine the their performance. IB 1 IB 3 IB 5 ID3 Nave Bayes MC4 Pitcher [0, 1, 2, 4, 11, 12] CV: 0. 43 T: 22.75 R: 116 [1, 2, 4, 7, 12, 15, 17, 37] CV: 0.43 T: 23.63 R: 118 [1, 2, 7, 12, 20, 37] CV: 0.58 T: 23.14 R: 115 1, 2, 7, 9, 11, 12, 19, 20, 37] CV: 1.30 T: 22.36 R: 115 [1, 2, 3, 6, 12, 18, 21, 30, 37] CV: 1.29 T: 17.60 R: 112 [1, 2, 10, 11, 15, 16, 17, 19, 20, 22, 23, 24, 25, 26, 27, 28, 30, 35, 38] CV: ....

[Article contains additional citation context not shown here]

D. Aha, editor, Lazy Learning, Kluwer Academic Publishers, The Netherlands, 1997.

....for various ratios of number of targets (n) to number of robots (m) problem include a very large search space, the need for communication or awareness of robot team members, and the diculty of assigning credit in an inherently cooperative problem. In this learning approach, lazy learning [1] is used to enable robot team members to build a memory of situation action pairs through random exploration of the CMOMMT problem. A reinforcement function gives the utility of a given situation. The pessimistic algorithm for each robot then uses the utility values to select the action that ....

....Markov Decision Process (POMDP) domain) the existence of local indicators that help individual robots perform their tasks, nor the use of symbolic representations. The following subsections describe this approach and its results in more detail. 5. 1 Lazy learning and Q learning Lazy learning [1] also called instance based learning promotes the principle of delaying the use of the gathered information until the necessity arises (see Fig. 4) The same pool of information (i.e. memory) is used for di erent behavior syntheses. The lazy memory provides a good way of reducing the duration ....

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

....schemes. In order for any optimal control technique to be successful, the structure of the system must be exploited in order to obtain tractable algorithms. As was demonstrated in [4] the types of problems described above may be cast as multidimensional system optimization; see [15] 8] and [3] for further details. The class of systems described in this paper fall under the class of spatially invariant systems. In [2] it is shown that these classes of problems may be tackled using transform techniques. The approach taken in [4] and further refined in this paper, is to consider the ....

N. K. Bose, editor. Multidimensional systems theory: progress, directions, and open problems in multidimensional systems. Kluwer Academic Publishers, 1985.

....constantly attracts the interest of many researchers both in theoretical foundations of computing and in application oriented fields. In the last two decades the fundamentals of data compression have been laid [3, 6, 9, 10, 15, 16, 17, 22] and efficiently applied to text and image compression [8, 11, 19]. Currently data compression is of increasing interest again because of the growing amount of data processed in applications and transferred over the internet. In particular compression of geometric data is currently an active research area [4, 12, 20, 21] and it is also important to perform ....

James A. Storer, editor. Image and Text Compression. Kluwer Academic Publishers, Norwell, Massachusetts, 1992.

....BiCMOS generally refers to CMOS BiCMOS where bipolar transistors are used to improve the driving capability of CMOS logic elements (Figure 1.1) In general this will improve Vdd Out In Figure 1.1: BiCMOS (BiNMOS) buffer. the driving capability of relatively long wires by about a factor of two [2] [22] A parallel multiplier does indeed have some long wires, and the long wires contribute significantly to the total delay, but the delay is not dominated by the long wires. A large number of short wires also contribute significantly to delay. The net effect is perhaps a 20 to 30 improvement ....

A. R. Alvarez, editor. BiCMOS Technology and Applications. Kluwer Academic Publishers, 1989.

....at each level in the order that it is performed: 1) region, 2) perceptual, 3) object part, 4) object and (5) scene. 4.2. 1 Region Region classification serves as an initial selection process similar to the one in [23] Current region classifiers are similar to IB1, a lazy learning algorithm [1] which is a variation of K Nearest Neighbor classification. Lazy learning algorithms perform classification by combining their stored data (i.e. the training set) when an input occurs. This makes it possible to have dynamic models that can be changed by the user over time since the examples ....

D. Aha, Editor. Lazy Learning. Kluwer Academic Publishers, The Netherlands, 1997.

....[24] synthesis system which implements the designs as a custom chip made using a set of communicating word parallel dedicated datapaths that are controlled by a central finitestate machine controller. HYPER generates the physical layout of the chips by using the LAGER silicon compilation system [46] at the backend datapaths are generated using the datapath compiler in LAGER, the FSM controller using the logic synthesis and tiling tools in LAGER, the datapath control logic using standard cells, and the overall chip using the macro cell place androute and pad ring generator tools that are ....

R. W. Brodersen, editor. Anatomy of a Silicon Compiler. Kluwer Academic Publishers, 1992.

....flexible compression system. 22 CODESIGN FOR REAL TIME VIDEO APPLICATIONS Table 2.4. Alternative realizations of the compression tasks. Compression Task Possible Realization Motion Estimation Hierarchical blockmatching [ 48] 20] phase correlation [ 48] Reduction of Wavelets [ 6] 39] 42] [ 115] [ 129] subband coding [ 211] Spatial Redundancy LOT [ 101] 130] 214] fractal compression [ 16] 98] 99] Entropy Coding Vector quantization [ 57] 63] 36] arithmetic coding [ 207] EMBEDDED VIDEO COMPRESSION SYSTEM The video compression algorithms described above require high ....

A. Laine, editor. Wavelet Theory and Application. Kluwer Academic Publishers, Dordrecht, Netherlands, 1993.

....is generally known as speedup learning, where the learner attempts to improve its own learning performance. The task of learning how to learn has cropped up repeatedly in the machine learning and artificial intelligence literature, culminating recently in the topic of multistrategy learning [30], where the learner is given multiple biases to choose from. The fundamental adversary being fought is the generality efficiency tradeoff: when a learner weakens its inductive bias and allows a greater variety of concepts to be learned, then it takes more examples, on average, to learn any ....

R. S. Michalski, editor. Multistrategy Learning. Kluwer Academic Publishers, 1993.

No context found.

T. (editor) Sasao. Logic Synthesis and Optimization. Kluwer Academic Publishers, Boston, Massachusetts, 1993. 130

....Briefly, the input signal is extended by folding at the boundaries in a symmetric or anti symmetric manner. The symmetry of the extension is determined by the length and symmetry characteristics of the filters and input signal, and furthermore by the constraint of perfect reconstruction [52, 10, 11]. The method of Symmetric Extension is used in the applications presented later. Figure 1.8 demonstrates symmetric and anti symmetric extensions of a signal. Boundary Wavelets It is possible [15] to explicitly construct boundary wavelets with as many vanishing moments as the original wavelet #, ....

....the two sides of this procedure are often referred to as a coder,orcodec (from CODerDECoder) 4.1.4 Quantization There are many different methods of quantization, of which only two are applied in this work. Introductory discussion of quantization and its place in lossy compression may be found in [48, 52]. Definition 4.4 (Quantization) Let X be a (possibly continuous) set, and K be a discrete set. Let Q and D be mappings Q : X K and D : K X. Q and D are such that #x D(Q(x) ###x D(d)##d # K Applying Q to some x X is called quantization, and Q(x) is the quantized value of x. Likewise, ....

[Article contains additional citation context not shown here]

P. N. Topiwala, editor. Wavelet Image and Video Compression. Kluwer Academic Publishers, 1998.

....environment for the visual description of an architecture should be available. Moreover, the tool environment should allow the user to perform syntactical checks on his model concerning e.g. the import export relations between the component interfaces. Architecture Description Languages (ADLs) [2] o er a well de ned set of notations for the description of an application architecture. Basically, an ADL allows the developer to describe the abstract organization of his system in terms of coarse grained architectural elements, independent of the element s implementation details. Apart from ....

....for our two level approach presented in this paper. The OCoN tool environment supports the visual development of OCoNs but not their relation to the architectural level. Other examples for visualizing architecture and (restricted forms of) their evolution can be found in ADL tool environments [2]. None of these tools allow a generic description of the visual model as we suggest in our approach using GenGEd. We model evolution steps by graph transformation which is also subject to considerable research. Software architecture recon guration based on graph transformation is presented by ....

P. Donohe, editor. Software Architecture. Kluwer Academic Publishers, 1999.

....of self contained software components is an important condition for the independent development of system parts (Open World Perspective) Newly developed components thus may be added to a system at di erent times within the design process or at run time. Architecture Description Languages (ADLs) Don99] o er a well de ned set of notations for the description of an application architecture following the decomposition paradigm mentioned above. Basically, an ADL allows the developer to describe the abstract organization of his system in terms of coarse grained architectural elements, independent ....

....nets than we adopted in our sample two level language. The OCoN tool environment supports the visual development of OCoNs but not their relation to the architectural level. Other examples for visualizing architecture and (restricted forms of) their evolution can be found in ADL tool environments [Don99] and [JR00] a tool for reengineering purposes. None of these tools allow a generic description of the visual model as we suggest in our approach using GenGEd. We model evolution steps by graph transformation which is also subject to considerable research. Software architecture recon guration ....

P. Donohe, editor. Software Architecture. Kluwer Academic Publishers, 1999.

....graphical models. 1.1 Introduction During the last few years, the use of probabilistic methods in artificial intelligence and machine learning has gained enormous popularity. In particular, probabilistic graphical models have become the preferred method for knowledge representation and reasoning [1]. The advantage of the probabilistic approach is that all assumptions are made explicit in the modeling process and that consequences, such as predictions on novel data, are assumption free and follow from a mechanistic computation. The drawback of the probabilistic approach is that the method is ....

M. Jordan, editor. Learning in Graphical Models. Kluwer Academic Publishers, 1996. NATO ASI Series.

....Although product line engineering is not yet widespread, we are beginning to have a better understanding of the processes, economics, and artifacts required to achieve the benefits of a product line approach. A number of case studies of product line successes have been published. For example, see [13]. Moreover, organizations such as the Software Engineering Institute are well on their way towards providing concrete guidelines and processes for the use of a product line approach [37] Like product line approaches, cross vendor integration standards require architectural frameworks that ....

P. Donohoe, editor. Software Architecture: TC2 First Working IFIP Conference on Software Architecture (WICSA1). Kluwer Academic Publishers, 1999.

No context found.

Michalski, R.S. (Ed.) (1993): Multistrategy Learning. Kluwer Academic Publishers.

No context found.

A. Oren, editor. Halophilic microogansims and their environments. Kluwer Academic Publishers, Dordrecht, 2001.

No context found.

M. Dror, editor. Arc Routing. Theory, Solutions and Applications. Kluwer Academic Publishers, Boston, MA, 2000.

No context found.

James A. Storer, editor. Image and Text Compression. Kluwer Academic Publishers, Norwell, Massachusetts, 1992.

No context found.

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

No context found.

D. Aha, editor. Lazy Learning. Kluwer Academic Publishers, 1997.

No context found.

P. Donohe, editor. Software Architecture. Kluwer Academic Publishers, 1999.

No context found.

M. Tomits, editor. Generalized LR Parsing. Kluwer Academic Publishers, 1991.

No context found.

J. A. Storer (Editor), Image and Text Compression, Kluwer Academic Publishers, USA, 1992.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC