606 research outputs found
A unifying framework for computational reinforcement learning theory:
No full textComputational learning theory studies mathematical models that allow one to formally analyze and compare the performance of supervised-learning algorithms such as their sample complexity. While existing models such as PAC (Probably Approximately Correct) have played an influential role in understanding the nature of supervised learning, they have not been as successful in reinforcement learning (RL). Here, the fundamental barrier is the need for active exploration in sequential decision problems.
An RL agent tries to maximize long-term utility by exploiting its knowledge about the problem, but this knowledge has to be acquired by the agent itself through exploring the problem that may reduce short-term utility. The need for active exploration is common in many problems in daily life, engineering, and sciences. For example, a Backgammon program strives to take good moves to maximize the probability of winning a game, but sometimes it may try novel and possibly harmful moves to discover how the opponent reacts in the hope of discovering a better game-playing strategy. It has been known since the early days of RL that a good tradeoff between exploration and exploitation is critical for the agent to learn fast (i.e., to reach near-optimal strategies with a small sample complexity), but a general theoretical analysis of this tradeoff remained open until recently.
In this dissertation, we introduce a novel computational learning model called KWIK (Knows What It Knows) that is designed particularly for its utility in analyzing learning problems like RL where active exploration can impact the training data the learner is exposed to. My thesis is that the KWIK learning model provides a flexible, modularized, and unifying way for creating and analyzing reinforcement-learning algorithms with provably efficient exploration. In particular, we show how the KWIK perspective can be used to unify the analysis of existing RL algorithms with polynomial sample complexity. It also facilitates the development of new algorithms with smaller sample complexity, which have demonstrated empirically faster learning speed in real-world problems. Furthermore, we provide an improved, matching sample complexity lower bound, which suggests the optimality (in a sense) of one of the KWIK-based algorithms known as delayed Q-learning.Ph.D.Includes bibliographical references (p. 238-261)by Lihong L
A STUDY OF TANG POETRY’S INFLUENCE ON AMERICAN POETRY: A NEW PERSPECTIVE OF “EIGHT BEAUTIES”
Get PDFTang poetry, the greatest treasure of classical Chinese poetry, has greatly influenced the American new poetry movement, which has been discussed and explained by scholars and experts in the academic world, but their researches are relatively macroscopic. Based on the micro-level "Eight Beauties" (beauty of form, musicality, image, emotion, diction, structure, allusion and gestalt) of “Harmony-Guided Three-Level Poetry Translation Criteria” proposed by Dr. Wang Feng in 2015, this paper aims to analyze how the eight beauties of Tang poetry were represented in American poems and exerted great influence on American poetry. Article visualizations
Family altruism and incentives
Get PDFThe author builds on the altruistic model of the family, to explore the strategic interaction between altruistic parents, and selfish children, when children's efforts are endogenous. If there is uncertainty about the amount of income the children will realize, and if parents have imperfect information, the children have an incentive to exert little effort, and to rely on their parent's altruistically motivated transfers. Because of this, parents face a tradeoff between the insurance that bequests implicitly provide their children, and the disincentive to work prompted by their altruism. The author shows that if parents can credibly commit to a pattern of transfers, they will choose not to compensate children in bad outcomes, as much as predicted by the standard (no uncertainty, no asymmetric information) dynastic model of the family. Alternatively, parents may choose to forgo any insurance, and offer a fixed level of bequest, to elicit greater effort from their children. The optimal transfers structure that the author derives, reconciles the predictions of the altruistic family model, with much of the existing evidence on inter-generational transfers, which suggests that parents compensate only partially, or not at all, for earnings differentials among their children. Moreover, the author shows that Ricardian equivalence holds in this setup, except when non-negativity constraints are binding.Economic Theory&Research,Environmental Economics&Policies,Health Economics&Finance,Educational Sciences,Safety Nets and Transfers
A clique-based discrete bat algorithm for influence maximization in identifying top-k influential nodes of social networks
No full textCompetition and Innovation: The Tango of the Market and Technology in the Competitive Landscape
No full textTechnology innovation can be a double-edged sword in helping a firm to address competitive pressures. We explore the relationships among market competition, technology competition, and firms\u27 advancement to a higher generation of production technology. Though market competition drives technology advancement as firms attempt to escape competition and technology competition also drives technology advancement as firms try to stay in the technology race, concurrent high market and technology competition lead a firm to defer advancement. We find supportive evidence with data on global flat panel display makers. Our findings shed light on how competition interacts with a firm\u27s technology advancement decision and, in general, a firm\u27s technology strategy. Author(s): Qian, Lihong; Wang, I. Kim Source: MANAGERIAL AND DECISION ECONOMICS, 38 (8): 1237-1247 DEC 2017 Document Type: Articl
Noninvasive photoacoustic microscopy of methemoglobin in vivo
Get PDFDue to the various causes of methemoglobinemia and its potential to be confused with other diseases, in vivo measurements of methemoglobin have significant applications in the clinic. Using photoacoustic microscopy (PAM), we quantified the average and the distributed percentage of methemoglobin both in vitro and in vivo. Based on the absorption spectra of methemoglobin, oxyhemoglobin, and deoxyhemoglobin, three wavelengths were chosen to differentiate methemoglobin from the others. The methemoglobin concentrations calculated from the photoacoustic signals agreed well with the preset concentrations. Then we imaged the methemoglobin percentage in microtubes that mimicked blood vessels. Average percentages calculated for five samples with different methemoglobin concentrations also agreed well with the preset values. Finally, we demonstrated the ability of PAM to detect methemoglobin in vivo in a mouse ear. Our results show that PAM can quantitatively image methemoglobin distribution in vivo
Photoacoustic Imaging and Spectroscopy [Book Review]
Get PDFPhotoacoustic Imaging and Spectroscopy is a multiauthored reference book that presents an advanced series of disparate chapters on the mathematical foundations, instrumentation, and applications of photoacoustic and thermoacoustic imaging. Lihong Wang, an eminent author, educator, scientist, and leader in the field of photoacoustic imaging and spectroscopy, is the editor of this book. Clearly this field is extremely active as evidenced by the diversity, the scope, and the quality of the field’s published literature. Nevertheless, I was surprised to read the back cover, which I think is overreaching with the claim that photoacoustics may make as dynamic a contribution to modern medicine as the discovery of the x ray once did. While this may be a “typical” overstatement by the publisher’s marketing team, it should be noted that within one year of the discovery of x rays there were more than one thousand publications related to the topic
Genetic algorithm for feedback-based wavefront shaping in optical imaging
No full textTraditional optical devices rely on light propagation along a straight path. However, when the light propagates through a blurred medium, its direction get scattered by microscopic particles. This inhomogeneous distortion results in a diffused focus point. Light scattering is one of the main limitations for the optical imaging. This limitation decreases the resolution in depth. Therefore, the ability of focusing light at a desired position has a huge worthwhile for applications of optical imaging. Over the past few years, it was shown that light can be focused inside an object even with strong scattering particles, just by shaping the wavefront of the incident beam. The most successful approaches for light focusing at the presence of scattering objects are feedback-based optical wavefront shaping. In this paper, an iterative feedback-based wavefront shaping is proposed. It uses the genetic algorithm. In summary, we aim to obtain a high intensity in the focus point with fewer steps in iteration while increase the signal-to-noise. The simulations results show that both the above mentioned goals are achieved using the proposed method.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.ImPhys/Acoustical Wavefield Imagin
Experimental demonstration of a multi-target detection technique using an X-band optically steered phased array radar
No full textCan reforming global institutions help developing countries share more in the benefits from globalization?
Get PDFGlobalization could significantly expand trade, international investment, and technological advances, but the gains from global integration have been unevenly distributed across and within nations. Greater global interdependence has also brought greater macroeconomic volatility, resulting in several serious financial crises in the second half of the 1990s. The global matrix of Bretton Woods and United Nations institutions that developed starting in the 1940s, formed under a different balance of power, in a world of fixed exchange rates and limited capital mobility. Since the 1960s regional financial institutions have emerged because of the greater autonomy of different regions and the greater financial needs of development. The author reviews different proposals for reform of the international financial institutions and changes in the roles of the International Monetary Fund (IMF) and the World Bank. He highlights the implications for developing countries of (1) Policy conditionality. (2) The countercyclical role of multilaterals'lending. (3) Greater lending to middle-income than to low-income developing countries. (3) Access to liquidity at times of crisis. (4) Mechanisms for giving low-income countries a greater voice in IMF and World Bank decisionmaking. The author streses the overlapping responsibilities of the Bretton Woods and regional financial institutions and the need to reassess the allocation of responsibilities and to develop better coordination mechanisms between these institutions. Those designing institutional reform must consider the corporate capabilities of each type of institution. The corporate cultures of global and regional institutions differ. So does the kind of knowledge they generate and disseminate, and so do patterns of interactions with, and mechanisms for representation of, client countries.Finally, the author calls attention to the need to harmonize national and global growth-oriented policies in a way that reduces volatility and promotes social equity.Environmental Economics&Policies,Governance Indicators,Financial Intermediation,Economic Theory&Research,Banks&Banking Reform
- …
