US 7,707,131 B2Grant
Thompson strategy based online reinforcement learning system for action selection
Issue Date:2010-04-27
•17 Claims
•6 Drawing Sheets
Abstract
A system and method for online reinforcement learning is provided. In particular, a method for performing the explore-vs.-exploit tradeoff is provided. Although the method is heuristic, it can be applied in a principled manner while simultaneously learning the parameters and/or structure of the model (e.g., Bayesian network model).
The system includes a model which receives an input (e.g., from a user) and provides a probability distribution associated with uncertainty regarding parameters of the model to a decision engine. The decision engine can determine whether to exploit the information known to it or to explore to obtain additional information based, at least in part, upon the explore-vs.-exploit tradeoff (e.g., Thompson strategy). A reinforcement learning component can obtain additional information (e.g., feedback from a user) and update parameter(s) and/or the structure of the model. The system can be employed in scenarios in which an influence diagram is used to make repeated decisions and maximization of long-term expected utility is desired.
Metadata
Assignee
- Microsoft Corporation
Inventors
- David M. Chickering
- Timothy S. Paek
- Eric J. Horvitz
Application Information
Application Number:US 11/169,503
Filing Date:2005-06-29
Priority Date:2005-03-08
Art Unit:2129
Classifications
IPC:
G06N5/04G06N7/00G06N7/02
Patent Drawings (6 sheets)
Description
Reference to Related Applications
[0001] This application claims the benefit of U.S. Provisional Application Ser. No. 60/659,689 filed on Mar. 8, 2005, and entitled SYSTEMS AND METHODS THAT FACILITATE ONLINE LEARNING FOR DIALOG SYSTEMS, the entirety of which is incorporated herein by reference.
Background
[0002] Influence diagrams can be employed to facilitate decision making. These models can be constructed by an expert (typically with the aid of a decision analyst) and/or learned from data. Generally, supervised learning techniques for “ordinary” Bayesian networks apply easily (e.g., with little modification) to learn the structure and parameters of an influence diagram. If an influence diagram is going to be used repeatedly to make decisions, it is desirable to use resulting observed data to improve the model over time.
[0003] Reinforcement learning deals with learning how to act in an environment. One of the central problems in reinforcement learning is deciding when to explore and when to exploit. In particular, given a current state of an environment and given a model about the expected (short-term) reward for performing each action in that state, the system can “exploit” by performing the action that has the highest expected reward. On the other hand, because the model may be uncertain about the environment, the system can instead choose to perform a sub-optimal short-term action in order to gain information and improve the model for the long term (“explore”).
Summary
[0004] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.
[0005] An online reinforcement learning system and method (e.g., adapting parameters of an influence diagram) is provided. Reinforcement learning deals with learning how to act in an environment. One of the central problems in reinforcement learning is deciding when to explore and when to exploit. In particular, given a current state of an environment and given a model about the expected (short-term) reward for performing each action in that state, the system can “exploit” by performing the action that has the highest expected reward. On the other hand, because the model may be uncertain about the environment, the system can instead choose to perform a sub-optimal short-term action in order to gain information and improve the model in the long term.
[0006] A particular method for performing the explore-vs.-exploit tradeoff can be employed. Although the method is heuristic, it can be applied in a principled manner while simultaneously learning the parameters and/or structure of the model (e.g., influence-diagram model).
[0007] For example, if an influence diagram is going to be used repeatedly to make decisions, the resulting observed data can be employed to improve the model over time. This can be important, for example, when the original parameter estimates are uncertain and/or when the domain itself is changing over time. If the decision maker is only interested in maximizing the immediate expected utility at each point in time, the maximum a posteriori (MAP) parameter values can be used and the resulting influence diagram can be solved. The parameters and/or structure of the influence diagram can be updated by performing incremental supervised learning techniques using the data observed after each decision.
[0008] On the other hand, if the decision maker is instead interested in maximizing long-term expected utility, the best action to take at any particular point in time can be sub-optimal in terms of immediate reward. The system includes a model which receives an input (e.g., from a user) and provides a probability distribution associated with uncertainty regarding parameters of the model to a decision engine. The decision engine can determine whether to exploit the information known to it or to explore to obtain additional information. A reinforcement learning component can obtain additional information (e.g., feedback from a user) and update parameter(s) and/or the structure of the model. The system can be employed in scenarios in which an influence diagram is used to make repeated decisions and maximization of long-term expected utility is desired.
[0009] The decision engine can employ a heuristic algorithm with the probability distribution to determine whether to exploit information known to the model or explore to gain additional information. The heuristic algorithm can be based, for example, upon a description by Thompson, infra, for selecting actions. In this example, the following action-selection strategy can be employed (the Thompson strategy): choose each decision policy stochastically, choosing policy Π with probability that Π is optimal.
[0010] The Thompson strategy is not optimal for any cost function (such as the sum of all values) except in degenerate problems; this follows immediately for problems in which there is a unique optimal action for every decision because the Thompson strategy chooses actions randomly. The strategy does, however, have three properties that make it particularly attractive. First, the strategy is intuitively appealing. Second, the strategy is easy to implement. Finally, as the posterior distributions over the parameters become more and more peaked, the strategy correctly converges to the optimal “exploit only” strategy.
[0011] In this example, the probability that a policy is optimal is equal to the total prior probability mass over the set of parameters that result in that policy being optimal. This means that by simply sampling parameters from their probability distribution and solving for the optimal policy, the decision engine can select policies based on the probability that they are optimal. Furthermore, in order to facilitate sampling from p(θ), the variable-specific parameter distributions can be modeled as Dirichlet or normal-Wishart distributions, which are the typical choices for discrete and continuous variables, respectively, when learning Bayesian networks.
[0012] To the accomplishment of the foregoing and related ends, certain illustrative aspects are described herein in connection with the following description and the annexed drawings. These aspects are indicative, however, of but a few of the various ways in which the principles of the claimed subject matter may be employed and the claimed subject matter is intended to include all such aspects and their equivalents. Other advantages and novel features of the claimed subject matter may become apparent from the following detailed description when considered in conjunction with the drawings.
Brief Description of the Drawings
[0013] FIG. 1 is a block diagram of an online reinforcement learning system.
[0014] FIG. 2 is an influence diagram.
[0015] FIG. 3 is an expanded influence diagram.
[0016] FIG. 4 is a flow chart of a method that facilitates online reinforcement learning.
[0017] FIG. 5 illustrates an example operating environment.
Detailed Description
[0018] The claimed subject matter is now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the claimed subject matter. It may be evident, however, that the claimed subject matter may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing the claimed subject matter.
[0019] As used in this application, the terms “component,” “handler,” “model,” “system,” and the like are intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. By way of illustration, both an application running on a server and the server can be a component. One or more components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. The components may communicate via local and/or remote processes such as in accordance with a signal having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, and/or across a network such as the Internet with other systems via the signal). Computer components can be stored, for example, on computer readable media including, but not limited to, an ASIC (application specific integrated circuit), CD (compact disc), DVD (digital video disk), ROM (read only memory), floppy disk, hard disk, EEPROM (electrically erasable programmable read only memory) and memory stick in accordance with the claimed subject matter.
[0020] An online reinforcement learning system and method (e.g., adapting parameters of an influence diagram) are provided. Adaptation of parameters of an influence diagram can be important, for example, when the influence diagram is used repeatedly to make decisions and there exists uncertainty regarding its parameters. Influence diagrams can be employed, for example, to facilitate repeated stock-trade decisions using an influence diagram. In this example, performance of the system is desired to be improved over time in order to maximize long-term return.
[0021] Referring to FIG. 1 , an online reinforcement learning system 100 is illustrated. Reinforcement learning deals with learning how to act in an environment. One of the central problems in reinforcement learning is deciding when to explore and when to exploit. In particular, given a current state of an environment and given a model 110 about the expected (short-term) reward for performing each action in that state, the system 100 can “exploit” by performing the action that has the highest expected reward. On the other hand, because the model 110 may not be completely accurate, the system 100 can instead choose to perform a sub-optimal (with respect to short term) action in order to gain information and improve the model 110 for the long term.
[0022] In one example, a particular method for performing this explore-vs.-exploit tradeoff is provided. Although the method is heuristic, it can be applied in a principled manner while simultaneously learning the parameters and/or structure of the model 110 (e.g., influence-diagram model).
[0023] For example, if an influence diagram is going to be used repeatedly to make decisions, the resulting observed data can be used to improve the model over time. This can be important, for example, when the original parameter estimates are uncertain and/or when the domain itself is changing over time. If the decision maker is only interested in maximizing the immediate expected utility at each point in time, the maximum a posteriori (MAP) parameter values can be used and the resulting influence diagram can be solved. The parameters and/or structure of the influence diagram can be updated by performing incremental supervised learning techniques using the data observed after each decision.
[0024] On the other hand, if the decision maker is instead interested in maximizing long-term expected utility, the best action to take at any particular point in time can be sub-optimal in terms of immediate reward. The system 100 includes a model 110 which receives an input (e.g., from a user) and provides a probability distribution associated with uncertainty regarding parameters of the model to a decision engine 120. The decision engine 120 can determine whether to exploit the information known to it or to explore to obtain additional information. A reinforcement learning component 130 can obtain additional information (e.g., feedback from a user) and update parameter(s) and/or the structure of the model 110.
[0025] The system 100 can be employed in scenarios in which an influence diagram is used to make repeated decisions and maximization of long-term expected utility is desired. The system 100 results in a model 110 which is an interesting hybrid of two specific types of models that are studied in the reinforcement-learning community. In particular, the problem to be solved by the system 100 is similar to the classic N-armed bandit problem, except that each “pull” of a bandit corresponds to solving a generalization of a finite-horizon Markov decision process (MDP) (e.g., the influence diagram). Accordingly, the problem to be solved by the system 100 can be referred to as the Bandit ID (Influence Diagram) problem. For all but the most simple (long-term) utility functions, and for all but the most simple influence diagrams, the problem of identifying the optimal set of actions to take is intractable due to the uncertainty in the parameter values.
[0026] Those skilled in the art will recognize that researchers in adaptive control, experimental design, and reinforcement learning have used numerous heuristic explore-vs.-exploit strategies to solve N-armed bandit problems. One such strategy, originally due to W. R. Thompson (W. R. Thompson, On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. Biometricka, 25:285-294 (1933)), is to choose actions stochastically based on the probability they are optimal (the Thompson strategy). This strategy has a number of appealing properties. First, it is easy to implement without approximations. Second, the strategy converges, in the limit of the number of observations in the environment, to the optimal policy. Finally, the approach is intuitively appealing as a reasonable strategy for interacting with the environment. In one example, the system 100 can apply the Thompson strategy to a Bandit ID problem.
[0027] Model 110
[0028] In this application, a variable is denoted by an upper case token (e.g., A, Θ) and a state or value of that variable is denoted by the same token in lower case (e.g., a, θ). Additionally, sets are denoted with bold-face capitalized tokens (e.g., A, Θ) and corresponding sets of values are denoted by bold-face lower case tokens (e.g., a, θ).
[0029] An influence diagram is a graphical model defined over a domain consisting of three types of variables: chance variables U, decision variables D, and value variables V. The influence diagram also contains a single utility function that is a deterministic function of all of the value variables. An influence diagram contains a set of parameters Θ that characterize the conditional distributions of the non-decision variables. In particular, the diagram defines the probability distribution p(U,V|D, Θ) via the local distributions stored within the non-decision nodes:
[0030]
where Pa(X) denotes the set of parents for node X, and where Θx denotes the subset of parameters in Θ that define the local distribution of X. Parents of a decision node D represent the nodes for which the values will be known at the time decision D is made.
where Pa(X) denotes the set of parents for node X, and where Θx denotes the subset of parameters in Θ that define the local distribution of X. Parents of a decision node D represent the nodes for which the values will be known at the time decision D is made.
[0031] Those skilled in the art will recognize that this definition of an influence diagram deviates from the traditional definition of an influence diagram because multiple value nodes are allowed that are not necessarily deterministic. The use of multiple stochastic value nodes is an optimization to allow efficient representation of a factored utility function (e.g., the utility is the sum of all the values nodes), and has been used by other researchers.
[0032] Referring briefly to FIG. 2 , an exemplary influence diagram 200 is illustrated. In this example, circular nodes 210 represent chance variables, square nodes 220 represent decision variables, and diamond nodes 230 represent value variables.
[0033] If the parameters Θ of an influence diagram are known with certainty, well-studied inference techniques can be applied to “solve” for the optimal sequence of decisions represented in that diagram. In particular, corresponding to each setting θ of the parameters is an optimal policy Π (Θ) that prescribes, for each decision node in the influence diagram, what the best choice is as a function of the values of the observed variables. In one example, the policy Π (Θ) is not constructed explicitly, but rather as each decision needs to be made, an inference algorithm is run to determine the best action to take.
[0034] In one example, it is assumed that the parameters of the influence diagram are not known with certainty, but rather a prior probability distribution p(Θ) exists. Briefly, learning the parameters of an influence from observed data is a straightforward extension to the problem of learning the parameters of a Bayesian network; the only significant difference is that the joint distribution of interest in an influence diagram is one that is conditioned on a set of decisions.
[0035] In this example, the standard parameter-independence assumption from the Bayesian-network-learning community is made:
[0036]
[0037] Note that the parameter-independence assumption implies that there is no “tying” of the parameters within an influence diagram. It is assumed that the influence diagram is going to be used repeatedly, and after each such decision instance, the parameter distribution is updated given the resulting observed values. Di, Ui, and Vi are used to denote the decision, chance, and value variables, respectively, in the ith decision instance. Dn=∪i =1n Di is used to denote the union of all decision variables up throughthe nth decision instance; Un and Vn is used to denote the corresponding unions for the chance and value variables, respectively. As in the case of learning Bayesian networks, it is assumed that the distribution p(U,V|D) is exchangeable, and thus all chance and value variables in the ith decision instance are independent of previous observations once Θ is known. Consequently, after n decision instances,
p(Θ|Un,Dn,Vn)
can be calculated and this distribution can be used to determine the optimal policy for the next decision instance. In particular, if it is assumed that all values are observed in each decision instance, the posterior distribution in decision instance n+1 can be calculated efficiently using the MAP parameter values {circumflex over (θ)} from the posterior:
p(Un+1, Vn+1|Dn+1,Un,Dn,Vn)=p(Un+1, Vn+1|Dn+1, {circumflex over (θ)})
p(Θ|Un,Dn,Vn)
can be calculated and this distribution can be used to determine the optimal policy for the next decision instance. In particular, if it is assumed that all values are observed in each decision instance, the posterior distribution in decision instance n+1 can be calculated efficiently using the MAP parameter values {circumflex over (θ)} from the posterior:
p(Un+1, Vn+1|Dn+1,Un,Dn,Vn)=p(Un+1, Vn+1|Dn+1, {circumflex over (θ)})
[0038] Thus, after n decision instances, if the MAP values of the parameter distribution can be computed, the (locally) optimal policy for decision instance n+1 can be identified by performing the well known inference algorithms under the assumption that the parameters are known to be equal to the MAP values.
[0039] Referring to FIG. 3 , an expanded influence diagram 300 is illustrated. In this example, as is common in the literature on learning Bayesian networks from data, the influence diagram 300 has been expanded to include all of the instance-specific variables and to include explicit nodes 310 to represent the unknown parameters. Not included in the Figure for the sake of simplicity are the “information arcs,” in which all chance nodes from previous time slices are assumed to be known or observed before making any current decision.
[0040] The N-armed bandit problem, a well-studied topic in many disciplines such as reinforcement-learning, is a simple example of the on-line influence-diagram learning problem. In the bandit problem, it is assumed that a person is gambling with n slot machines, where the ith slot machine pays a dollar with probability θi. The person is uncertain about the payoff parameters, so it is desirable to construct a policy that, given n total pulls, maximizes the person's expected return. This problem exemplifies the classic “explore vs. exploit” problem that planning algorithms face. In particular, if the person is only interested in maximizing expected return on the next pull, it is easy to see that the optimal policy is to “exploit” the person's knowledge and play the machine with the highest expected return (e.g., play the machine for which the expected value of the payoff parameter is highest). Given k>1 remaining pulls, however, it may be better to “explore” by experimenting with a machine that may not have the highest expected return in order to gain more information about the corresponding payoff parameter; the idea is that by gaining information about the parameter values, the best machines to play in the long run can be identified with more certainty.
[0041] Assuming a beta prior for each θi, if the number of pulls is bounded by k, the n-armed bandit problem can be solved exactly, assuming a desire to maximize the total payoff, using dynamic-programming techniques. When there are an infinite number of pulls remaining, the problem can be solved exactly in a similar manner, but only if it is desired to maximize a discounted total payoff(e.g., a dollar k turns from now is valued as δk dollars now). Even when exact methods are possible, oftentimes closed form solutions may not be feasible, so researchers often apply heuristic methods for solving these problems. Heuristic methods can also be preferred when the parameter values are changing over time; this corresponds to a violation of the exchangeability assumption described in the previous section.
[0042] The problem to be solved by the system 100 is a generalization of the N-armed bandit problem: each “pull” corresponds to implementing a strategy in an influence diagram-bandit ID problem. Just as in the n-armed-bandit problem, if the total return over multiple decision instances is of interest, the explore-vs-exploit dilemma is explored. Unfortunately, solving this dilemma exactly is infeasible in general, so heuristic or approximate algorithms are designed.
[0043] To summarize, based, at least in part, upon information received from a user, the model 110 can provide a probability distribution associated with uncertainty regarding parameters to the decision engine 120. The decision engine 120 can employ a heuristic algorithm with the probability distribution to determine whether to exploit information known to the model 110 or explore to gain additional information.
[0044] The heuristic algorithm can be based, for example, upon a description by Thompson (1933), supra, for selecting actions. Thompson (1933) was interested in how to allocate patients between the control arm and the treatment arm of a clinical trial. Thompson (1933) argued that in order to minimize a particular measure of risk, the fraction of patients assigned to the treatment group should be precisely the probability that the treatment arm of the clinical trial was optimal (e.g., the probability that the treatment would result in a better outcome for a patient). Thompson (1933) then derived a recursive formula for the probability that the treatment arm is optimal, assuming a Beta distribution over the treatment and control probabilities of success. In a later published article (Thompson, W. R. (1935), On the theory of apportionment, American Journal of Math, 57:450-456), Thompson generalized the derivation to the problem where there are n possible treatments.
[0045] The decision engine 120 can employ this heuristic in order to select actions. There is clearly a strong connection between the problem Thompson studied and the n-arm bandit problem; the main difference is that Thompson was concerned with allocating all plays ahead of time, whereas in the bandit problem, the parameter distribution can be updated after each play. In this example, the following action-selection strategy can be employed (the Thompson strategy): choose each decision policy stochastically, choosing policy Π with probability that Π optimal.
[0046] The Thompson strategy is not optimal for any cost function (such as the sum of all values) except in degenerate problems; this follows immediately for problems in which there is a unique optimal action for every decision because the Thompson strategy chooses actions randomly. The strategy does, however, have three properties that make it particularly attractive. First, the strategy is intuitively appealing. Second, in one example, the strategy can be implemented exactly without any approximation. Finally, as the posterior distributions over the parameters become more and more peaked, the strategy correctly converges to the optimal “exploit only” strategy.
[0047] Researchers in the reinforcement learning field have applied the Thompson strategy to a variety of bandit problems. Similar to Thompson (1933), the researchers have also derived the probability that the payoff for each machine is optimal assuming a Beta parameter prior. Thus, this research showed that the Thompson strategy is superior to many other simple explore-vs.-exploit strategies found in the reinforcement-learning literature.
[0048] Yet other researchers have applied the Thompson strategy for action selection in a model-free reinforcement-learning problem, using a Bayesian approach to Q-learning. These researchers recognized that although deriving the probability that each decision is optimal may be difficult, sampling decisions according to this distribution is simple. Unlike the bandit problems considered by other researchers, these researchers found that the Thompson strategy did not perform particularly well compared to some alternative strategies.
[0049] With regard to implementation of the Thompson strategy, recall that if the parameter values θ of an influence diagram are known, optimal performance at each step can be achieved by simply performing the unique optimal decision policy Π (θ) in each decision instance. It follows that the probability any particular policy Π optimal is:
Popt(Π)=∫9s.t.Π(θ)=ΠP(θ)
Popt(Π)=∫9s.t.Π(θ)=ΠP(θ)
[0050] In other words, the probability that a policy is optimal is equal to the total prior probability mass over the set of parameters that result in that policy being optimal. This means that by simply sampling parameters from their probability distribution and solving for the optimal policy, the decision engine 120 can select policies based on the probability that they are optimal. Furthermore, in order to facilitate sampling from p(θ), the variable-specific parameter distributions can be modeled as Dirichlet or normal-Wishart distributions, which are the typical choices for discrete and continuous variables, respectively, when learning Bayesian networks.
[0051] It can be observed that in the limit of the number of decision instances, the Thompson strategy for policy selection converges to the optimal policy. In particular, any policy with non-zero probability will eventually be tried a sufficient number of times to deem that it is or is not the best policy.
[0052] An unfortunate property of the Thompson strategy—and of almost all the standard explore-vs.-exploit strategies in the literature—is the fact that the strategy does not use the number of remaining decision instances in its choice for the next action. If there is only one more decision instance, it would be preferred to use the MAP parameter values as opposed to a random sample. Thus, in one example, the decision engine 120 can artificially decrease the variance of the parameter distributions as the last decision instance is approached (e.g., using the Dirichlet and normal-Wishart distributions).
[0053] As discussed previously, heuristic strategies such as the Thompson strategy are often applied even when an optimal strategy can be computed because the parameters may be changing over time. In one example, non-stationary parameters can be a problem for the Thompson strategy if the variances of the parameter distributions get too low. This can be mitigated by artificially keeping these variances large enough that exploration continues so that the model can adapt.
[0054] It is to be appreciated that the system 100, the model 110, the decision engine 120 and/or the reinforcement learning component 130 can be computer components as that term is defined herein.
[0055] Turning briefly to FIG. 4 , a methodology that may be implemented in accordance with the claimed subject matter is illustrated. While, for purposes of simplicity of explanation, the methodology is shown and described as a series of blocks, it is to be understood and appreciated that the claimed subject matter is not limited by the order of the blocks, as some blocks may, in accordance with the claimed subject matter, occur in different orders and/or concurrently with other blocks from that shown and described herein. Moreover, not all illustrated blocks may be required to implement the methodology.
[0056] The claimed subject matter may be described in the general context of computer-executable instructions, such as program modules, executed by one or more components. Generally, program modules include routines, programs, objects, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically the functionality of the program modules may be combined or distributed as desired in various embodiments.
[0057] Referring to FIG. 4 , a method that facilitates online reinforcement learning 400 is illustrated. At 410, an input is received, for example, from a user. At 420, a probability distribution associated with uncertainty regarding parameters of a model is determined. At 430, using the probability distribution, a heuristic technique (e.g., Thompson strategy) is employed to select an action. At 440, parameter(s) of the model are updated based, at least in part, upon feedback associated with the selected action.
[0058] In order to provide additional context for various aspects of the claimed subject matter, FIG. 5 and the following discussion are intended to provide a brief, general description of a suitable operating environment 510 in which various aspects of the claimed subject matter may be implemented. While the claimed subject matter is described in the general context of computer-executable instructions, such as program modules, executed by one or more computers or other devices, those skilled in the art will recognize that the claimed subject matter can also be implemented in combination with other program modules and/or as a combination of hardware and software. Generally, however, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular data types. The operating environment 510 is only one example of a suitable operating environment and is not intended to suggest any limitation as to the scope of use or functionality of the claimed subject matter. Other well known computer systems, environments, and/or configurations that may be suitable for use with the claimed subject matter include but are not limited to, personal computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include the above systems or devices, and the like.
[0059] With reference to FIG. 5 , an exemplary environment 510 includes a computer 512. The computer 512 includes a processing unit 514, a system memory 516, and a system bus 518. The system bus 518 couples system components including, but not limited to, the system memory 516 to the processing unit 514. The processing unit 514 can be any of various available processors. Dual microprocessors and other multiprocessor architectures also can be employed as the processing unit 514.
[0060] The system bus 518 can be any of several types of bus structure(s) including the memory bus or memory controller, a peripheral bus or external bus, and/or a local bus using any variety of available bus architectures including, but not limited to, an 8-bit bus, Industrial Standard Architecture (ISA), Micro-Channel Architecture (MSA), Extended ISA (EISA), Intelligent Drive Electronics (IDE), VESA Local Bus (VLB), Peripheral Component Interconnect (PCI), Universal Serial Bus (USB), Advanced Graphics Port (AGP), Personal Computer Memory Card International Association bus (PCMCIA), and Small Computer Systems Interface (SCSI).
[0061] The system memory 516 includes volatile memory 520 and nonvolatile memory 522. The basic input/output system (BIOS), containing the basic routines to transfer information between elements within the computer 512, such as during start-up, is stored in nonvolatile memory 522. By way of illustration, and not limitation, nonvolatile memory 522 can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory 520 includes random access memory (RAM), which acts as external cache memory. By way of illustration and not limitation, RAM is available in many forms such as synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM).
[0062] Computer 512 also includes removable/nonremovable, volatile/nonvolatile computer storage media. FIG. 5 illustrates, for example, a disk storage 524. Disk storage 524 includes, but is not limited to, devices like a magnetic disk drive, floppy disk drive, tape drive, Jaz drive, Zip drive, LS-100 drive, flash memory card, or memory stick. In addition, disk storage 524 can include storage media separately or in combination with other storage media including, but not limited to, an optical disk drive such as a compact disk ROM device (CD-ROM), CD recordable drive (CD-R Drive), CD rewritable drive (CD-RW Drive) or a digital versatile disk ROM drive (DVD-ROM). To facilitate connection of the disk storage devices 524 to the system bus 518, a removable or non-removable interface is typically used such as interface 526.
[0063] It is to be appreciated that FIG. 5 describes software that acts as an intermediary between users and the basic computer resources described in suitable operating environment 510. Such software includes an operating system 528. Operating system 528, which can be stored on disk storage 524, acts to control and allocate resources of the computer system 512. System applications 530 take advantage of the management of resources by operating system 528 through program modules 532 and program data 534 stored either in system memory 516 or on disk storage 524. It is to be appreciated that the claimed subject matter can be implemented with various operating systems or combinations of operating systems.
[0064] A user enters commands or information into the computer 512 through input device(s) 536. Input devices 536 include, but are not limited to, a pointing device such as a mouse, trackball, stylus, touch pad, keyboard, microphone, joystick, game pad, satellite dish, scanner, TV tuner card, digital camera, digital video camera, web camera, and the like. These and other input devices connect to the processing unit 514 through the system bus 518 via interface port(s) 538. Interface port(s) 538 include, for example, a serial port, a parallel port, a game port, and a universal serial bus (USB). Output device(s) 540 use some of the same type of ports as input device(s) 536. Thus, for example, a USB port may be used to provide input to computer 512, and to output information from computer 512 to an output device 540. Output adapter 542 is provided to illustrate that there are some output devices 540 like monitors, speakers, and printers among other output devices 540 that require special adapters. The output adapters 542 include, by way of illustration and not limitation, video and sound cards that provide a means of connection between the output device 540 and the system bus 518. It should be noted that other devices and/or systems of devices provide both input and output capabilities such as remote computer(s) 544.
[0065] Computer 512 can operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 544. The remote computer(s) 544 can be a personal computer, a server, a router, a network PC, a workstation, a microprocessor based appliance, a peer device or other common network node and the like, and typically includes many or all of the elements described relative to computer 512. For purposes of brevity, only a memory storage device 546 is illustrated with remote computer(s) 544. Remote computer(s) 544 is logically connected to computer 512 through a network interface 548 and then physically connected via communication connection 550. Network interface 548 encompasses communication networks such as local-area networks (LAN) and wide-area networks (WAN). LAN technologies include Fiber Distributed Data Interface (FDDI), Copper Distributed Data Interface (CDDI), Ethernet/IEEE 802.3, Token Ring/IEEE 802.5 and the like. WAN technologies include, but are not limited to, point-to-point links, circuit switching networks like Integrated Services Digital Networks (ISDN) and variations thereon, packet switching networks, and Digital Subscriber Lines (DSL).
[0066] Communication connection(s) 550 refers to the hardware/software employed to connect the network interface 548 to the bus 518. While communication connection 550 is shown for illustrative clarity inside computer 512, it can also be external to computer 512. The hardware/software necessary for connection to the network interface 548 includes, for exemplary purposes only, internal and external technologies such as, modems including regular telephone grade modems, cable modems and DSL modems, ISDN adapters, and Ethernet cards.
[0067] What has been described above includes examples of the claimed subject matter. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the claimed subject matter, but one of ordinary skill in the art may recognize that many further combinations and permutations of the claimed subject matter are possible. Accordingly, the claimed subject matter is intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims. Furthermore, to the extent that the term “includes” is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term “comprising” as “comprising” is interpreted when employed as a transitional word in a claim.
Claims
What is claimed is:
1. An online reinforcement learning system comprising components embodied on a computer readable storage medium, the components when executed by one or more processors, update a model based upon reinforcement learning, the components comprising:
a model comprising an influence diagram with at least one chance node, the model receiving an input and providing a probability distribution associated with uncertainty regarding parameters of the model;
a decision engine that selects an action based, at least in part, upon the probability distribution, the decision engine employing a Thompson strategy heuristic technique to maximize long term expected utility when selecting the action, wherein the decision engine decreases a variance of a distribution of the parameters as a last decision instance is approached; and
a computer-implemented reinforcement learning component that modifies at least one of the parameters of the model based upon feedback associated with the selected action, the parameters defining distributions over discrete variables and continuous variables, uncertainty of the parameters expressed using Dirichlet priors for conditional distributions of discrete variables of the model, and, Normal-Wishart priors for distributions of continuous variables of the model, wherein the modified model is stored.
2. The system of claim 1, used when the parameters of the model are changing over time.
3. The system of claim 1, wherein the decision engine employs a maximum a posterior of the parameters when there is only one more decision instance remaining.
4. The system of claim 1, wherein the decision engine artificially increases the variance of a distribution of the parameters.
5. The system of claim 1, wherein the computer-implemented reinforcement learning component further modifies the structure of the model based, at least in part, upon the feedback associated with the selected action.
6. The system of claim 1, wherein the feedback comprises an input from a user of the system.
7. The system of claim 6, wherein the input from the user comprises a verbal utterance.
8. The system of claim 1, wherein the feedback comprises a lack of an input from a user of the system in a threshold period of time.
9. The system of claim 1, where one or more parameters of the model change over a period of time.
10. The system of claim 1, the parameters defining distributions over variables, where the variables comprise chance variables, decision variables and/or value variables.
11. The system of claim 1, employed repeatedly to facilitate decision making.
12. The system of claim 11, wherein the parameter(s) are updated prior to a next repetition.
13. The system of claim 1, the model comprising a Markov decision process represented as an Influence diagram.
14. The system of claim 1 employed as part of a dialog system.
15. An online reinforcement learning method comprising:
determining a probability distribution associated with uncertainty regarding parameters of a model, the model comprising an influence diagram with at least one chance node;
employing a computer-implemented Thompson strategy heuristic technique to select an action based, at least in part, upon the probability distribution, wherein a variance of a distribution of the parameters is artificially increased to be large enough that the model continues to adapt;
updating at least one parameter of the model based, at least in part, upon feedback associated with the selected action, the parameters defining distributions over discrete variables and continuous variables, uncertainty of the parameters expressed using Dirichlet priors for conditional distributions of discrete variables of the model, and, Normal-Wishart priors for distributions of continuous variables of the model; and
storing the updated model on a computer readable storage medium.
16. The method of claim 15, wherein the feedback comprises an input from a user or a lack of an input from the user in a threshold period of time.
17. A computer readable medium having stored thereon computer executable instructions for carrying out the method of claim 15.
Patent Citations (109)
| Patent | Date | Inventor | Cited By |
|---|---|---|---|
| US4228496(A) | 1980-10-01 | Katzman et al. | Applicant |
| US4769771(A) | 1988-09-01 | Lippmann et al. | Applicant |
| US5165018(A) | 1992-11-01 | Simor | Applicant |
| US5274821(A) | 1993-12-01 | Rouquie | Applicant |
| US5428792(A) | 1995-06-01 | Conner et al. | Applicant |
| US5428803(A) | 1995-06-01 | Chen et al. | Applicant |
| US5452447(A) | 1995-09-01 | Nelson et al. | Applicant |
| US5493692(A) | 1996-02-01 | Theimer et al. | Applicant |
| US5544321(A) | 1996-08-01 | Theimer et al. | Applicant |
| US5555376(A) | 1996-09-01 | Theimer et al. | Applicant |
| US5566302(A) | 1996-10-01 | Khalidi et al. | Applicant |
| US5577251(A) | 1996-11-01 | Hamilton et al. | Applicant |
| US5603054(A) | 1997-02-01 | Theimer et al. | Applicant |
| US5611050(A) | 1997-03-01 | Theimer et al. | Applicant |
| US5621809(A) | 1997-04-01 | Bellegarda et al. | Applicant |
| US5664173(A) | 1997-09-01 | Fast | Applicant |
| US5732274(A) | 1998-03-01 | O'Neill | Applicant |
| US5787251(A) | 1998-07-01 | Hamilton et al. | Applicant |
| US5806060(A) | 1998-09-01 | Borgida et al. | Applicant |
| US5812865(A) | 1998-09-01 | Theimer et al. | Applicant |
| US5826269(A) | 1998-10-01 | Hussey | Applicant |
| US5864810(A) | 1999-01-01 | Digalakis et al. | Applicant |
| US5899997(A) | 1999-05-01 | Ellacott | Applicant |
| US5950196(A) | 1999-09-01 | Pyreddy et al. | Applicant |
| US5961582(A) | 1999-10-01 | Gaines | Applicant |
| US5991754(A) | 1999-11-01 | Raitto et al. | Applicant |
| US5995959(A) | 1999-11-01 | Friedman et al. | Applicant |
| US6032144(A) | 2000-02-01 | Srivastava et al. | Applicant |
| US6128612(A) | 2000-10-01 | Brereton et al. | Applicant |
| US6144999(A) | 2000-11-01 | Khalidi et al. | Applicant |
| US6163856(A) | 2000-12-01 | Dion et al. | Applicant |
| US6173266(B1) | 2001-01-01 | Marx et al. | Applicant |
| US6182061(B1) | 2001-01-01 | Matsuzawa et al. | Applicant |
| US6253181(B1) | 2001-06-01 | Junqua | Applicant |
| US6389393(B1) | 2002-05-01 | Gong | Applicant |
| US6466232(B1) | 2002-10-01 | Newell et al. | Applicant |
| US6513046(B1) | 2003-01-01 | Abbott, III et al. | Applicant |
| US6516322(B1) | 2003-02-01 | Meredith | Applicant |
| US6523026(B1) | 2003-02-01 | Gillis | Applicant |
| US6523027(B1) | 2003-02-01 | Underwood | Applicant |
| US6549915(B2) | 2003-04-01 | Abbott, III et al. | Applicant |
| US6556960(B1) | 2003-04-01 | Bishop et al. | Examiner |
| US6584459(B1) | 2003-06-01 | Chang et al. | Applicant |
| US6598169(B1) | 2003-07-01 | Warwick et al. | Applicant |
| US6615322(B2) | 2003-09-01 | Arimilli et al. | Applicant |
| US6654734(B1) | 2003-11-01 | Mani et al. | Applicant |
| US6691148(B1) | 2004-02-01 | Zinky et al. | Applicant |
| US6747675(B1) | 2004-06-01 | Abbott et al. | Applicant |
| US6766330(B1) | 2004-07-01 | Chen et al. | Applicant |
| US6766348(B1) | 2004-07-01 | Combs et al. | Applicant |
| US6791580(B1) | 2004-09-01 | Abbott et al. | Applicant |
| US6799162(B1) | 2004-09-01 | Goronzy et al. | Applicant |
| US6801223(B1) | 2004-10-01 | Abbott et al. | Applicant |
| US6812937(B1) | 2004-11-01 | Abbott et al. | Applicant |
| US6826555(B2) | 2004-11-01 | Nunez | Applicant |
| US6826562(B1) | 2004-11-01 | Leung et al. | Applicant |
| US6842877(B2) | 2005-01-01 | Robarts et al. | Applicant |
| US6845507(B2) | 2005-01-01 | Kenton | Applicant |
| US6910068(B2) | 2005-06-01 | Zintel et al. | Applicant |
| US6912529(B1) | 2005-06-01 | Kolfman | Applicant |
| US6944184(B1) | 2005-09-01 | Miller et al. | Applicant |
| US6950821(B2) | 2005-09-01 | Faybishenko et al. | Applicant |
| US7055142(B2) | 2006-05-01 | Meredith et al. | Applicant |
| US7089294(B1) | 2006-08-01 | Baskey et al. | Applicant |
| US7117487(B2) | 2006-10-01 | Meredith et al. | Applicant |
| US7143167(B2) | 2006-11-01 | Kampe et al. | Applicant |
| US7292976(B1) | 2007-11-01 | Hakkani-Tur et al. | Applicant |
| US2001/0040590(A1) | 2001-11-01 | Abbott et al. | Applicant |
| US2001/0040591(A1) | 2001-11-01 | Abbott et al. | Applicant |
| US2001/0043231(A1) | 2001-11-01 | Abbott et al. | Applicant |
| US2001/0043232(A1) | 2001-11-01 | Abbott et al. | Applicant |
| US2002/0032689(A1) | 2002-03-01 | Abbott, III et al. | Applicant |
| US2002/0044152(A1) | 2002-04-01 | Abbott, III et al. | Applicant |
| US2002/0052930(A1) | 2002-05-01 | Abbott et al. | Applicant |
| US2002/0052963(A1) | 2002-05-01 | Abbott et al. | Applicant |
| US2002/0054130(A1) | 2002-05-01 | Abbott, III et al. | Applicant |
| US2002/0054174(A1) | 2002-05-01 | Abbott et al. | Applicant |
| US2002/0078204(A1) | 2002-06-01 | Newell et al. | Applicant |
| US2002/0080155(A1) | 2002-06-01 | Abbott et al. | Applicant |
| US2002/0080156(A1) | 2002-06-01 | Abbott et al. | Applicant |
| US2002/0083025(A1) | 2002-06-01 | Robarts et al. | Applicant |
| US2002/0083158(A1) | 2002-06-01 | Abbott et al. | Applicant |
| US2002/0087525(A1) | 2002-07-01 | Abbott et al. | Applicant |
| US2002/0099817(A1) | 2002-07-01 | Abbott et al. | Applicant |
| US2002/0143944(A1) | 2002-10-01 | Traversat et al. | Applicant |
| US2002/0147771(A1) | 2002-10-01 | Traversat et al. | Applicant |
| US2002/0169852(A1) | 2002-11-01 | Schaeck | Applicant |
| US2002/0184070(A1) | 2002-12-01 | Chen et al. | Applicant |
| US2003/0018694(A1) | 2003-01-01 | Chen et al. | Applicant |
| US2003/0046401(A1) | 2003-03-01 | Abbott et al. | Applicant |
| US2003/0058277(A1) | 2003-03-01 | Bowman-Amuah | Applicant |
| US2003/0126136(A1) | 2003-07-01 | Omoigui | Applicant |
| US2003/0154476(A1) | 2003-08-01 | Abbott, III et al. | Applicant |
| US2003/0163316(A1) | 2003-08-01 | Addison et al. | Applicant |
| US2003/0163461(A1) | 2003-08-01 | Gudbjartsson et al. | Applicant |
| US2003/0212540(A1) | 2003-11-01 | Meredith et al. | Applicant |
| US2003/0212671(A1) | 2003-11-01 | Meredith et al. | Applicant |
| US2003/0212672(A1) | 2003-11-01 | Meredith et al. | Applicant |
| US2003/0212761(A1) | 2003-11-01 | Meredith et al. | Applicant |
| US2004/0148363(A1) | 2004-07-01 | Hadi | Applicant |
| US2004/0181390(A1) | 2004-09-01 | Manson | Applicant |
| US2005/0033582(A1) | 2005-02-01 | Gadd et al. | Applicant |
| US2005/0034078(A1) | 2005-02-01 | Abbott et al. | Applicant |
| US2005/0125232(A1) | 2005-06-01 | Gadd | Applicant |
| US2006/0058999(A1) | 2006-03-01 | Barker et al. | Applicant |
| US2006/0195321(A1) | 2006-08-01 | Deligne et al. | Applicant |
| US2008/0059188(A1) | 2008-03-01 | Konopka et al. | Applicant |
| EP713257(A1) | 1996-05-01 | Applicant | |
| WO9800787 | 1998-01-01 | Applicant |
Non-Patent Literature (85)
- Walker, Marilyn, “An Application of Reinforcement Learning to Dialogue Strategy Selection in a Spoken Dialogue System for Email”, pp. 387-416, 2000.Examiner
- Stone, Matthew, “Foundations of Language Interaction”, pp. 1-4, 2001.Examiner
- Paek et al., Tim, “On the Utility of Decision-Theoretic Hidden Subdialog”, 2003.Examiner
- Horvitz et al., Eric, “Harnessing Models of Users' Goals to Mediate Clarification Dialog in Spoken Language Systems”, 2001.Examiner
- Singh et al, Satinder, “Reinforcement Learning for Spoken Dialgoue Systems”, 2000.Examiner
- Stone, Matthew. “Foundations of Language Interaction”, pp. 1-4, 2001.Examiner
- U.S. Appl. No. 11/170,584, filed Jun. 29, 2005, Paek et al.Applicant
- U.S. Appl. No. 11/170,998, filed Jun. 29, 2005, Paek et al.Applicant
- U.S. Appl. No. 11/170,999, filed Jun. 29, 2005, Paek et al.Applicant
- Andy Harter, et al., A Distributed Location System for the Active Office, IEEE Network, 1994, pp. 62-70.Applicant
- Guanling Chen, et al., A Survey of Context-Aware Mobile Computing Research, Dartmouth Computer Science Technical Report, 2000, 16 pages.Applicant
- William Noah Schilt, A System Architecture for Context-Aware Mobile Computing, Columbia University, 1995, 153 pages.Applicant
- Mike Spreitzer, et al., Providing Location Information in a Ubiquitous Computing Environment, SIGOPS '93, 1993, pp. 270-283.Applicant
- Marvin Theimer, et al., Operating System Issues for PDAs, In Fourth Workshop on Workstation Operating Systems, 1993, 7 pages.Applicant
- Roy Want, Active Badges and Personal Interactive Computing Objects, IEEE Transactions on Consumer Electronics, 1992, 11 pages, vol. 38—No. 1.Applicant
- Bill N. Schilit, et al., The ParcTab Mobile Computing System, IEEE WWOS-IV, 1993, 4 pages.Applicant
- Bill Schilit, et al., Context-Aware Computing Applications, In Proceedings of the Workshop on Mobile Computing Systems and Applications, Dec. 1994. pp. 85-90.Applicant
- Bill N. Schilit, et al., Customizing Mobile Applications, Proceedings Usenix Symposium on Mobile and Location Independent Computing, Aug. 1993, 9 pages.Applicant
- Mike Spreitzer, et al., Architectural Considerations for Scalable, Secure, Mobile Computing with Location Information, In The 14th International Conference on Distributed Computing Systems, Jun. 1994, pp. 29-38.Applicant
- Mike Spreitzer et al., Scalable, Secure, Mobile Computing with Location Information, Communications of the ACM, Jul. 1993, 1 page, vol. 36—No. 7.Applicant
- Roy Want, et al., The Active Badge Location System, ACM Transactions on Information Systems, Jan. 1992, pp 91-102, vol. 10—No. 1.Applicant
- Mark Weiser, Some Computer Science Issues in Ubiquitous Computing, Communications of the ACM, Jul. 1993, pp. 75-84, vol. 36—No. 7.Applicant
- M. Billinghurst, et al., An Evaluation of Wearable Information Spaces, Proceedings of the Virtual Reality Annual International Symposium, 1998, 8 pages.Applicant
- Bradley J. Rhodes, Remembrance Agent: A continuously running automated information retrieval system, The Proceedings of The First International Conference on The Practical Application Of Intelligent Agents and Multi Agent Technology, 1996, pp. 487-495.Applicant
- Eric Horvitz, et al., In Pursuit of Effective Handsfree Decision Support: Coupling Bayesian Inference, Speech Understanding, and User Models, 1995, 8 pages.Applicant
- Bradley J. Rhodes, The Wearable Remembrance Agent: A System for Augmented Theory, The Proceedings of The First International Symposium on Wearable Computers, Oct. 1997, pp. 123-128.Applicant
- Eric Horvitz, et al., Attention-Sensitive Alerting in Computing Systems, Microsoft Research, Aug. 1999.Applicant
- Bill N. Schilit, et al., Disseminationg Active Map Information to Mobile Hosts, IEEE Network, 1994, pp. 22-32, vol. 8—No. 5.Applicant
- Mark Billinghurst, et al., Wearable Devices: New Ways to Manage Information, IEEE Computer Society, Jan. 1999, pp. 57-64.Applicant
- Thad Eugene Starner, Wearable Computing and Contextual Awareness, Massachusetts Institute of Technology, Jun. 1999, 248 pages.Applicant
- Bradley J. Rhodes, The Wearable Remembrance Agent: A System for Augmented Memory, Personal Technologies Journal Special Issue on Wearable Computing, 1997, 12 pages.Applicant
- Workshop on Wearable Computing Systems, Aug. 19-21, 1996.Applicant
- Mark Billinghurst, Research Directions in Wearable Computing, University of Washington, May 1998,48 pages.Applicant
- Mark Weiser, The Computer for the 21st Century, Scientific American, Sep. 1991, 8 pages.Applicant
- Lee, Wayne. “Decision Theory and Human Behavior.” (1971) John Wiley & Sons, Inc. New York.Applicant
- Boutilier, C., Dean T., and Hanks, S. “Decision-theoretic planning: Structural assupmtions and computatonal leverage.” Journal of Artificial Intelligence Research, (1999). pp. 1-93, vol. 1:1.Applicant
- Cooper, G. F. “A method for using belief networks as influence diagrams.” Proceedings of the Ninth Conference on Uncertainty in Artificial Intelligence, Washington, D.C. (1993). pp. 55-63.Applicant
- Dearden, R., Friedman, N., and Russell, S. “Bayesian Q-learning.” Proceedings of the Fifteenth National Conference on Artificial Intelligence, (1998). pp. 761-768.Applicant
- Heckerman, D. “A Bayesian approach for learning casual networks.” Proceddings of the Eleventh Conference on Uncertainity in Artificial Intelligence, Montreal. (1995).Applicant
- Heckerman, D. “A tutorial on learning Bayesian networks.” Technical Report MSR-TR-95-06, Microsoft Research, Mar. 1995 (Revised Nov. 1996).Applicant
- Kaelbling, L. P., Littman, M. L. And Morre, A. W. “Reinforcment learning: A survey.” Journal of Articial Intelligence Research, (1996). pp. 237-285, vol. 4.Applicant
- Lauritzen, S. L. And Nilsson, D. “Representing and solving decisions problems with limited information.” Management Science, (2001). pp. 1235-1251, vol. 47.Applicant
- Singh, S., Litman, D., Kearns, M., and Walker, M. “Optimizing Dialogue managemant with reinforcement learning: Experiements with the nj-fun system.” Journal of Artificial Intelligence Research, (2002). pp. 105-133. vol. 16.Applicant
- Tatman, J. A. And Shachter, R.D. “Dynamic programming and influence diagrams.” IEEE Transactions on Sysytems, Man and Cybernetics, (1990). pp. 365-379. vol. 20.Applicant
- Thompson, W. R. “On The Liklihood That One Unknown Probability Exceeds Another In View of the Evidence of Two Samples.” Biometricka, (1933). pp. 285-294.Applicant
- Wyatt, J. PhD thesis, University of Edinburgh. Exploration and Inference in Learning from Reinforcement, (1997).Applicant
- Young, S. “Problilistic Methods in Spoken Dialogue Systems.” Philosophical Transactions of the Royal Society (Series A), (2000). pp. 1389-1402.Applicant
- Thompson, W. R. “On the theory of apportionment” American Journal of Math, (1935). vol. 57, pp. 285-294.Applicant
- T. Joachims, Text categorization with support vector machines: learning with many relevant features, Machine Learning, European Conference on Machine Learning, Apr. 21, 1998, pp. 137-142.Applicant
- International Search Report dated Sep. 29, 2003 for PCT Application Serial No. 00/20685, 3 Pages.Applicant
- Robert M. Losee, Jr., Minimizing information overload: the ranking of electronic messages, Journal of Information Science 15, Elsevier Science Publishers B.V., 1989, pp. 179-189.Applicant
- U.S. Appl. No. 11/169,503, filed Jun. 29, 2005, Chickering et al.Applicant
- OA Dated Feb. 11, 2009 for U.S. Appl. No. 11/170,998, 57 pages.Applicant
- OA Dated Feb. 19, 2009 for U.S. Appl. No. 11/170,999, 42 pages.Applicant
- Non-Final Office Action in U.S. Appl. No. 11/170,584 dated Dec. 24, 2008, 32 pages.Applicant
- Abramsky, “Computational Interpretations of Linear Logic” Theoretical Computer Science 111(1-2), 1993, 53 pgs.Applicant
- Berger, et al., “Sequentiality and the pi.-Calculus” Abramsky, S. (ed.) Proceedings: Typed Lambda Calculi and Applications, 5th Int'l Conference. Krakow, Poland, May 2-5, 2001, 56 pgs.Applicant
- Bray, et al., “Extensible Markup Language (XML) 1.0 (Second Edition)” Oct. 6, 2000, <<http://www.w3.org/ TR/20000/REC-xml-20001006.html>>, 50 pgs.Applicant
- Chaudhuri, et al., “A Robust, Optimization-Based Approach for Approximate Answering of Agregate Queries”, ACM, May 2001, 12 pgs.Applicant
- Cleaveland, et al., “An Operational Framework for Value-Passing Processes” Feb. 1994, Proceedings of the 21st ACM SIGPLAN-SIGACT symposium on Principles of programming languages, pp. 326-338.Applicant
- Emmerich, “Implementing Incremental Code Migration with XML”, ACM Press, Proc. 22nd Int. Conf. on Software Engineering, 2000, pp. 397-406, 10 pgs.Applicant
- Feijs, “Modeling Microsoft COM Using pi.-Calculus”, Lecture Notes in Computer Science, vol. 1709, 1999, pp. 1343-1363, 31 pgs.Applicant
- Fournet, et al., “A Calculus of Mobile Agents” 1996, <<http://citeseer.ist.psu.edu/fournet96calculus.html>>, Last accessed Sep. 7, 2007, 16 pgs.Applicant
- Fournet, et al., “The join calculus: a language for distributed mobile programming” Sep. 2000 <<http://citeseer.ist.psu.edu/432452.html>>, Last accessed Sep. 7, 2007, 36 pgs.Applicant
- Haas, et al., “Extensible Query Processing in Starburst” ACM, Jun. 1989, 12 pgs.Applicant
- Honda, et al., “A Uniform Type Structure for Secure Information Flow”, Jan. 2002 ACM SIGPLAN Notices, Proceedings of the 29th ACM SIGPLAN-SIGACT symposium on Principles of programming languages POPL '02, vol. 37, Issue 1, 12 pgs.Applicant
- Lee, et al., “On Modeling Real-time Mobile Processes”, Jan. 2002, Australian Computer Science Commuications, Proceedings of the twenty-fifth Australiasian conference on Computer science—vol. 4 CRPITS '02, vol. 24 Issue 1, pp. 139-147.Applicant
- Mehta, “BizTalk Server 2000 Business Process Orchestration” 2001, <<http://sites.computer.org/debull/AO1mar/mehta.ps.>> Last accessed Sep. 7, 2007, 6 pgs.Applicant
- Microsoft, “What's New in the Apr. 2001 Microsoft XML Parser 4.0 Technology Preview” Apr. 2001, <<http://msdn.microsoft.com/archive>>, 3 pgs.Applicant
- Milner, “The Polyadic pi.-Calculus: A Tutorial” Proceedings: Int'l Summer School on Logic Algebra of Specification. Marktoberdorf, Germany, 1992, 50 pgs.Applicant
- Mohr, “Introduction to BizTalk Server 2000”, 2000, <<http://www.topxml.com/conference/wrox/2000—vegas/text/stephen—biztalk.pdf>>, Last accessed Sep. 7, 2007, 10 pgs.Applicant
- Parson “Using Java Reflection to Automate Extension Language Parsing” IEEE. Dec. 1999, 14 pgs.Applicant
- Peim, et al., “Query Processing with Description Logic Ontologies Over Object-Wrapped Databases” IEEE, 2002, 10 pgs.Applicant
- Pericas-Geertsen, “XML-Fluent Mobile Agents” Dec. 2001, <<http://citeseer.ist.psu.edu/pericas-geertsen99xmlfluent.html>>, Last accessed Sep. 7, 2007, 90 pgs.Applicant
- Rasche, et al., “Configurable Services for Mobile Users”, Proc. Seventh IEEE International Workshop on Object-Oriented Real-Time Dependable Systems, 2002, 17 pgs.Applicant
- Riely, et al., “A Typed Language for Distributed Mobile Proceses (Extended abstract)”, Jan. 1998 Proceedings of the 25th ACM SIGPLAN-SIGACT symposium on Principles of programming languages, pp. 378-390.Applicant
- Thatte, “XLANG, Web Services for Business Process Design” 2001, <<http://web.archive.org/web/20020207122217/ http://www.gotdotnet.com/team/xmlwsspecs/xlang-c/default.htm>>, Last accessed Sep. 2, 2008, 35 pgs.Applicant
- Uramoto, et al.,“Composition of Web Services Based on Their Behavioral Descriptions” Journal of the Japanese Society for Artificial Intelligence, vol. 17, No. 4. The Japanese Society for Artificial Intelligence, Japan, Jul. 1, 2002, 10 pgs.Applicant
- Yoshida, et al. “Strong Normalisation in the pi.-Calculus”. MCS Technical Report, Sep. 2001, Mar. 2001, 22 pgs.Applicant
- U.S. Appl. No. 11/169,503, filed on Jun. 29, 2005, Chickering, et al., “Action selection for reinforcement learning using influence diagrams”.Applicant
- U.S. Appl. No. 11/170,584, filed on Jun. 29, 2005, Pack, et al., “Easy generation and automatic training of spoken dialog systems using text-to-speech ”.Applicant
- U.S. Appl. No. 11/170,999, filed on Jun. 29, 2005, Paek, et al., “Online learning for dialog systems”.Applicant
- OA Dated Dec. 24, 2008 for U.S. Appl. No. 11/170,584, p. 32.Applicant
- OA Dated Jan. 9, 2009 for U.S. Appl. No. 11/169,503, 14 Pages.Applicant
- OA Dated Oct. 28, 2009 for U.S. Appl. No. 11/169,503, 11 Pages.Applicant