Physical Sciences and Mathematics Commons^™

When Will Information Retrieval Be “Good Enough”?, James Allan

James Allan

We describe a user study that examined the relationship between the quality of an Information Retrieval system and the e#11;ectiveness of its users in performing a task. The task involves #12;nding answer facets of questions pertaining to a collection of newswire documents over a six month period. We arti#12;cially created sets of ranked lists at increasing levels of quality by blending the output of a state-of-the-art retrieval system with truth data created by annotators. Subjects performed the task by using these ranked lists to guide their labeling of answer passages in the retrieved articles. We found that as system accuracy …

Hard Track Overview In Trec 2004 High Accuracy Retrieval From Documents, James Allan

James Allan

The HARD track of TREC 2004 aims to improve the accuracy of information retrieval through the use of three techniques: (1) query metadata that better describes the information need, (2) focused and time-limited interaction with the searcher through \clari#12;cation forms", and (3) incorporation of passage-level relevance judgments and retrieval. Participation in all three aspects of the track was excellent this year with about 10 groups trying something in each area. No group was able to achieve huge gains in e#11;ectiveness using these techniques, but some improvements were found and enthusiasm for the clari#12;cation forms (in particular) remains high. The track …

An Intrinsic Reward Mechanism For Efficient Exploration, Özgür Şimşek, Andrew G. Barto

Andrew G. Barto

How should a reinforcement learning agent act if its sole purpose is to efficiently learn an optimal policy for later use? In other words, how should it explore, to be able to exploit later? We formulate this problem as a Markov Decision Process by explicitly modeling the internal state of the agent and propose a principled heuristic for its solution. We present experimental results in a number of domains, also exploring the algorithm’s use for learning a policy for a skill given its reward function—an important but neglected component of skill discovery.

Basic-Block Instruction Scheduling Using Reinforcement Learning And Rollouts, Amy Mcgovern, Eliot Moss, Andrew G. Barto

Andrew G. Barto

The execution order of a block of computer instructions on a pipelined machine can make a difference in its running time by a factor of two or more. In order to achieve the best possible speed, compilers use heuristic schedulers appropriate to each specific architecture implementation. However, these heuristic schedulers are time-consuming and expensive to build. We present empirical results using both rollouts and reinforcement learning to construct heuristics for scheduling basic blocks. In simulation, both the rollout scheduler and the reinforcement learning scheduler outperformed a commercial scheduler on several applications.

Associative Search Network: A Reinforcement Learning Associative Memory, Andrew Barto, Richard Sutton, Peter Brouwer

Andrew G. Barto

An associative memory system is presented which does not require a "teacher" to provide the desired associations. For each input key it conducts a search for the output pattern which optimizes an external payoff or reinforcement signal. The associative search network (ASN) combines pattern recognition and function optimization capabilities in a simple and effective way. We define the associative search problem, discuss conditions under which the associative search network is capable of solving it, and present results from computer simulations. The synthesis of sensory-motor control surfaces is discussed as an example of the associative search problem.

Automatic Discovery Of Subgoals In Reinforcement Learning Using Diverse Density, Amy Mcgovern, Andrew G. Barto

Andrew G. Barto

This paper presents a method by which a reinforcement learning agent can automatically discover certain types of subgoals online. By creating useful new subgoals while learning, the agent is able to accelerate learning on the current task and to transfer its expertise to other, related tasks through the reuse of its ability to attain subgoals. The agent discovers subgoals based on commonalities across multiple paths to a solution. We cast the task of finding these commonalities as a multiple-instance learning problem and use the concept of diverse density to find solutions. We illustrate this approach using several gridworld tasks.

Adaptive Critics And The Basal Ganglia, Andrew G. Barto

Andrew G. Barto

No abstract provided.

Accelerating Reinforcement Learning Through The Discovery Of Useful Subgoals, Amy Mcgovern, Andrew G. Barto

Andrew G. Barto

An ability to adjust to changing environments and unforeseen circumstances is likely to be an important component of a successful autonomous space robot. This paper shows how to augment reinforcement learning algorithms with a method for automatically discovering certain types of subgoals online. By creating useful new subgoals while learning, the agent is able to accelerate learning on a current task and to transfer its expertise to related tasks through the reuse of its ability to attain subgoals. Subgoals are created based on commonalities across multiple paths to a solution. We cast the task of finding these commonalities as a …

Intrinsically Motivated Reinforcement Learning: A Promising Framework For Developmental Robot Learning, Andrew Stout, George D. Konidaris, Andrew G. Barto

Andrew G. Barto

One of the primary challenges of developmental robotics is the question of how to learn and represent increasingly complex behavior in a self-motivated, open-ended way. Barto, Singh, and Chentanez (Barto, Singh, & Chentanez 2004; Singh, Barto, & Chentanez 2004) have recently presented an algorithm for intrinsically motivated reinforcement learning that strives to achieve broad competence in an environment in a tasknonspecific manner by incorporating internal reward to build a hierarchical collection of skills. This paper suggests that with its emphasis on task-general, self-motivated, and hierarchical learning, intrinsically motivated reinforcement learning is an obvious choice for organizing behavior in developmental robotics. …

Socially Guided Machine Learning, Andrea Lockerd Thomaz, Cynthia Breazeal, Andrew G. Barto, Rosalind Picard

Andrew G. Barto

Social interaction will be key to enabling robots and machines in general to learn new tasks from ordinary people (not experts in robotics or machine learning). Everyday people who need to teach their machines new things will find it natural for to rely on their interpersonal interaction skills. This thesis provides several contributions towards the understanding of this Socially GuidedMachine Learning scenario. While the topic of human input to machine learning algorithms has been explored to some extent, prior works have not gone far enough to understand what people will try to communicate when teaching a machine and how algorithms …

Identifying Useful Subgoals In Reinforcement Learning By Local Graph Partitioning, Özgür Şimşek, Alicia Wolfe, Andrew Barto

Andrew G. Barto

We present a new subgoal-based method for automatically creating useful skills in reinforcement learning. Our method identifies subgoals by partitioning local state transition graphs—those that are constructed using only the most recent experiences of the agent. The local scope of our subgoal discovery method allows it to successfully identify the type of subgoals we seek—states that lie between two densely-connected regions of the state space—while producing an algorithm with low computational cost.

Using Relative Novelty To Identify Useful Temporal Abstractions In Reinforcement Learning, Özgür Şimşek, Andrew G. Barto

Andrew G. Barto

We present a new method for automatically creating useful temporal abstractions in reinforcement learning. We argue that states that allow the agent to transition to a different region of the state space are useful subgoals, and propose a method for identifying them using the concept of relative novelty. When such a state is identified, a temporallyextended activity (e.g., an option) is generated that takes the agent efficiently to this state. We illustrate the utility of the method in a number of tasks.

Linear Least-Squares Algorithms For Temporal Difference Learning, Steven Bradtke, Andrew Barto

Andrew G. Barto

We introduce two new temporal difference (TD) algorithms based on the theory of linear leastsquares function approximation. We define an algorithm we call Least-Squares TD (LS TD) for which we prove probability-one convergence when it is used with a function approximator linear in the adjustable parameters. We then define a recursive version of this algorithm, Recursive Least-Squares TD (RLS TD). Although these new TD algorithms require more computation per time-step than do Sutton's TD(A) algorithms, they are more efficient in a statistical sense because they extract more information from training experiences. We describe a simulation experiment showing the substantial improvement …

Scheduling Straight-Line Code Using Reinforcement Learning And Rollouts, Amy Mcgovern, Eliot Moss, Andrew G. Barto

Andrew G. Barto

The execution order of a block of computer instructions on a pipelined machine can make a difference in its running time by a factor of two or more. In order to achieve the best possible speed, compilers use heuristic schedulers appropriate to each specific architecture implementation. However, these heuristic schedulers are time-consuming and expensive to build. We present empirical results using both rollouts and reinforcement learning to construct heuristics for scheduling basic blocks. In simulation, the rollout scheduler outperformed a commercial scheduler, and the reinforcement learning scheduler performed almost as well as the commercial scheduler.

Rapid Landscape Transformation In South Island, New Zealand, Following Initial Polynesian Settlement, David B. Mcwethy, Cathy Whitlock, Janet M. Wilmshurst, Matt S. Mcglone, Mairie Fromont, Xun Li, Ann Dieffenbacher-Krall, William O. Hobbs, Sherilyn C. Fritz, Edward R. Cook

Department of Earth and Atmospheric Sciences: Faculty Publications

Humans have altered natural patterns of fire for millennia, but the impact of human-set fires is thought to have been slight in wet closed-canopy forests. In the South Island of New Zealand, Polynesians (Māori), who arrived 700–800 calibrated years (cal y) ago, and then Europeans, who settled ∼150 cal y ago, used fire as a tool for forest clearance, but the structure and environmental consequences of these fires are poorly understood. High-resolution charcoal and pollen records from 16 lakes were analyzed to reconstruct the fire and vegetation history of the last 1,000 y. Diatom, chironomid, and element concentration data were …

The Fierce Green Fire: Vol. 1 Issue 7, Wofford College Environmental Studies Program

The Fierce Green Fire

No abstract provided.

Biomarkers Of Mild Cognitive Impairment And Alzheimer's Disease, Mark A. Lovell, Bert C. Lynn

Chemistry Faculty Patents

A method for quantifying a neurodegenerative disorder in a patient that includes obtaining a fluid sample from the subject; measuring a protein biomarker complex in said fluid sample and correlating the measurement with mild cognitive impairment or Alzheimer's disease status. The biomarkers include those that comprise at least one of a transthyretin protein and/or a prostaglandin-H2 D-isomerase protein, and at least one second, different protein selected from a transthyretin, prostaglandin-H2 D-isomerase, beta-2-microglobulin, cystatin C, superoxide dismutase [Cu—Zn], plasma retinol-binding protein, phosphatidylethanolamine-binding protein, carbonic anhydrase 2, prostaglandin-H2 D-isomerase, and/or serotransferrin protein.

Precatalysts Useful In Polyolefin Polymerization Reactions, Omofolami Tesileem Ladipo, Richard Eaves, Alexey Zazybin, Sean Parkin

Chemistry Faculty Patents

Compounds are provided that are useful as precatlysts in the polymerization of olefins such as ethylene and propylene. Other compounds are useful as intermediates in the production of such precatalysts.

Transit Spectrophotometry Of The Exoplanet Hd 189733b Ii. New Spitzer Observations At 3.6 Μm, Jean-Michel Désert, David K. Sing, Alfred Vidal-Madjar, Guillaume Hébrard, David Ehrenreich, Alain Lecavelier Des Etangs, Vivien Parmentier, Roger Ferlet, Gregory W. Henry

Information Systems and Engineering Management Research Publications

Context. We present a new primary transit observation of the hot-jupiter HD 189733b, obtained at 3.6 μm with the Infrared Array Camera (IRAC) onboard the Spitzer Space Telescope. Previous measurements at 3.6 microns suffered from strong systematics, and conclusions could hardly be obtained with confidence on the water detection by comparison of the 3.6 and 5.8 microns observations.

Aims. We aim at constraining the atmospheric structure and composition of the planet and improving previously derived parameters.

Methods. We use a high-S/NSpitzer photometric transit light curve to improve the precision of the near infrared radius of …

On A Generalized Time-Varying Seir Epidemic Model With Mixed Point And Distributed Time-Varying Delays And Combined Regular And Impulsive Vaccination Controls, Ravi P. Agarwal, Manuel De La Sen, Asier Ibeas, Santiago Alonso-Quesada

Mathematics and System Engineering Faculty Publications

This paper discusses a generalized time-varying SEIR propagation disease model subject to delays which potentially involves mixed regular and impulsive vaccination rules. The model takes also into account the natural population growing and the mortality associated to the disease, and the potential presence of disease endemic thresholds for both the infected and infectious population dynamics as well as the lost of immunity of newborns. The presence of outsider infectious is also considered. It is assumed that there is a finite number of time-varying distributed delays in the susceptible-infected coupling dynamics influencing the susceptible and infected differential equations. It is also …

A Framework For Group Modeling In Agent-Based Pedestrian Crowd Simulations, Fasheng Qiu

Computer Science Dissertations

Pedestrian crowd simulation explores crowd behaviors in virtual environments. It is extensively studied in many areas, such as safety and civil engineering, transportation, social science, entertainment industry and so on. As a common phenomenon in pedestrian crowds, grouping can play important roles in crowd behaviors. To achieve more realistic simulations, it is important to support group modeling in crowd behaviors. Nevertheless, group modeling is still an open and challenging problem. The influence of groups on the dynamics of crowd movement has not been incorporated into most existing crowd models because of the complexity nature of social groups. This research develops …

Dynamic Data Driven Application System For Wildfire Spread Simulation, Feng Gu

Computer Science Dissertations

Wildfires have significant impact on both ecosystems and human society. To effectively manage wildfires, simulation models are used to study and predict wildfire spread. The accuracy of wildfire spread simulations depends on many factors, including GIS data, fuel data, weather data, and high-fidelity wildfire behavior models. Unfortunately, due to the dynamic and complex nature of wildfire, it is impractical to obtain all these data with no error. Therefore, predictions from the simulation model will be different from what it is in a real wildfire. Without assimilating data from the real wildfire and dynamically adjusting the simulation, the difference between the …

Superhydrophobic Thin Films Fabricated By Reactive Layer-By-Layer Assembly Of Azlactone-Functionalized Polymers, Maren E. Buck, Sarina C. Schwartz, David M. Lynn

Chemistry: Faculty Publications

We report an approach to the fabrication of superhydrophobic thin films that is based on the "reactive" layer-by-layer assembly of azlactone-containing polymer multilayers. We demonstrate that films fabricated from alternating layers of the azlactone functionalized polymer poly(2-vinyl- 4,4-dimethylazlactone) (PVDMA) and poly(ethyleneimine) (PEI) exhibit micro- and nanoscale surface features that result in water contact angles in excess of 150°. Our results reveal that the formation of these surface features is (i) dependent upon film thickness (i.e., the number of layers of PEI and PVDMA deposited) and (ii) that it is influenced strongly by the presence (or absence) of cyclic azlactone-functionalized oligomers …

Pauli Spin Blockade And Lifetime-Enhanced Transport In A Si/Sige Double Quantum Dot, C. B. Simmons, Teck Seng Koh, Nakul Shaji, Madhu Thalakulam, L. J. Klein, Hua Qin, H. Luo, D. E. Savage, M. G. Lagally, A. J. Rimberg

Dartmouth Scholarship

We analyze electron-transport data through a Si/SiGe double quantum dot in terms of spin blockade and lifetime-enhanced transport (LET), which is transport through excited states that is enabled by long spin-relaxation times. We present a series of low-bias voltage measurements showing the sudden appearance of a strong tail of current that we argue is an unambiguous signature of LET appearing when the bias voltage becomes greater than the singlet-triplet splitting for the (2,0) electron state. We present eight independent data sets, four in the forward-bias (spin-blockade) regime and four in the reverse-bias (lifetime-enhanced transport) regime and show that all eight …

Pre-Eruption Pressure, Temperature And Volatile Content Of Rhyolite Magma From The 1650 Ad Eruption Of Kolumbo Submarine Volcano, Greece, K. Cantner, S. Carey, H. Sigurdsson, G. Vougioukalakis, P. Nomikou, C. Roman, K. Bell, M. Alexandri

Christopher N. Roman

Biotite-bearing, crystal-poor rhyolite magma was the predominant magma type discharged during the 1650 AD explosive eruption of Kolumbo submarine volcano, Greece. The eruption produced thick sequences of pumice deposits (~100 m) in the upper crater walls of the volcano, but also led to the formation of extensive pumice rafts that were dispersed throughout the southern Aegean Sea, and subaerial tephra fallout as far east as Turkey. Preliminary estimates of pre-eruption volatile contents have been determined using the volatile-by-difference method on plagioclase-hosted melt inclusions and yield an average value of 6.0 wt.%. This corresponds to a pre-eruption storage pressure of 180 …

Do Cosmological Perturbations Have Zero Mean?, Christian Armendariz-Picon

Physics - All Scholarship

A central assumption in our analysis of cosmic structure is that cosmological perturbations have zero ensemble mean. This property is one of the consequences of statistically homogeneity, the invariance of correlation functions under spatial translations. In this article we explore whether cosmological perturbations indeed have zero mean, and thus test one aspect of statistical homogeneity. We carry out a classical test of the zero mean hypothesis against a class of alternatives in which perturbations have non-vanishing means, but homogeneous and isotropic covariances. Apart from Gaussianity, our test does not make any additional assumptions about the nature of the perturbations and …

Isomer-Dependent Adsorption And Release Of Cis- And Trans-Platin Anticancer Drugs By Mesopomus Silica Nanoparticles, Zhimin Tao, Youwei Xie, Jerry Goodisman, Tewodros Asefa

Chemistry - All Scholarship

We report on adsorption and release of the anticancer drugs cisplatin and transplatin from mesoporous silica nanomaterials, emphasizing the differences between cisplatin and its much less toxic isomer. Two types of particles, M.CM-41 and SBA-IS, were used, either as just synthesized or after calcination to remove the templates. The particles were characterized by TEM, nitrogen physisorption, and elemental analysis. The UV-vis spectra of cisplatin and transplatin were obtained and the intensities of several bands (205-210 nm, 210-220 nm, 220-235 nm, and 300-330 nm) were found proportional to drug concentrations, allowing their use for measuring drug concentration. To evaluate drug adsorption …

Uncertainty Associated With Modeling The Global Ionosphere, Janelle V. Jenniges, Ariel O. Acebal, Larry Gardner, Robert W. Schunk, Lie Zhu

Physics Student Research

A study has been conducted of the effect that different physical assumptions have on global models of the electron density distribution. The study was conducted with the Ionosphere Forecast Model (IFM) and the Ionosphere Plasmasphere Model (IPM) developed by Utah State University. Both physics-based, time-dependent, global models use the same empirical models for the neutral atmosphere (MSIS) and neutral wind (Horizontal Wind Model, HWM), but the altitude range, thermal structure, number of ion species, and magnetic 2ield are different. The IFM covers the altitude range from 90-1400 km, calculates the densities for four ions (NO+, O2+, N2+, O+), has a …

Dihomotopic Deadlock Detection Via Progress Shell Decomposition, David A. Cape, Stephen C. Jackson, Bruce M. Mcmillin

Computer Science Faculty Research & Creative Works

The classical problem of deadlock detection for concurrent programs has traditionally been accomplished by symbolic methods or by search of a state transition system. This work examines an approach that uses geometric semantics involving the topological notion of dihomotopy to partition the state space into components, followed by search of a reduced state space. Prior work partitioned the state-space inductively. in this work, a decomposition technique motivated by recursion coupled with a search guided by the decomposition is shown to effectively reduce the size of state transition systems. the reduced state space yields asymptotic improvement in overall runtime for verification. …

Paranets: A Parallel Network Architecture For The Future Internet, Khaled Harras, Abderrahmen Mtibaa

Computer Science Faculty Works

The evolution of networking technologies and portable devices has led users to expect connectivity anytime and everywhere. We have reached the point of seeing networking occur underwater, via aerial devices, and across space. While researchers push the true boundaries of networking to serve a wide range of environments, there is the challenge of providing robust network connectivity beyond the boundaries of the core internet, defined by fiber optics and well-organized backbones. As the internet edges expand, the expectation is that connectivity will be as good, in terms of high bandwidth and minimal interruption, as anywhere in the core. Such an …