Showing papers by "Carnegie Mellon University published in 2013"

emcee: The MCMC Hammer

Abstract: We introduce a stable, well tested Python implementation of the affine-invariant ensemble sampler for Markov chain Monte Carlo (MCMC) proposed by Goodman & Weare (2010). The code is open source and has already been used in several published projects in the astrophysics literature. The algorithm behind emcee has several advantages over traditional MCMC sampling methods and it has excellent performance as measured by the autocorrelation time (or function calls per independent sample). One major advantage of the algorithm is that it requires hand-tuning of only 1 or 2 parameters compared to ~N2 for a traditional algorithm in an N-dimensional parameter space. In this document, we describe the algorithm and the details of our implementation. Exploiting the parallelism of the ensemble method, emcee permits any user to take advantage of multiple CPU cores without extra effort. The code is available online at http://dan.iel.fm/emcee under the GNU General Public License v2.

Reinforcement learning in robotics: A survey

Abstract: Reinforcement learning offers to robotics a framework and set of tools for the design of sophisticated and hard-to-engineer behaviors. Conversely, the challenges of robotic problems provide both inspiration, impact, and validation for developments in reinforcement learning. The relationship between disciplines has sufficient promise to be likened to that between physics and mathematics. In this article, we attempt to strengthen the links between the two research communities by providing a survey of work in reinforcement learning for behavior generation in robots. We highlight both key challenges in robot reinforcement learning as well as notable successes. We discuss how contributions tamed the complexity of the domain and study the role of algorithms, representations, and prior knowledge in achieving these successes. As a result, a particular focus of our paper lies on the choice between model-based and model-free as well as between value-function-based and policy-search methods. By analyzing a simple problem in some detail we demonstrate how reinforcement learning approaches may be profitably applied, and we note throughout open questions and the tremendous potential for future research.

Stochastic variational inference

Abstract: We develop stochastic variational inference, a scalable algorithm for approximating posterior distributions. We develop this technique for a large class of probabilistic models and we demonstrate it with two probabilistic topic models, latent Dirichlet allocation and the hierarchical Dirichlet process topic model. Using stochastic variational inference, we analyze several large collections of documents: 300K articles from Nature, 1.8M articles from The New York Times, and 3.8M articles from Wikipedia. Stochastic inference can easily handle data sets of this size and outperforms traditional variational inference, which can only handle a smaller subset. (We also show that the Bayesian nonparametric topic model outperforms its parametric counterpart.) Stochastic variational inference lets us apply complex Bayesian models to massive data sets.

The State of US Health, 1990-2010: Burden of Diseases, Injuries, and Risk Factors

Abstract: Importance Understanding the major health problems in the United States and how they are changing over time is critical for informing national health policy. Objectives To measure the burden of diseases, injuries, and leading risk factors in the United States from 1990 to 2010 and to compare these measurements with those of the 34 countries in the Organisation for Economic Co-operation and Development (OECD) countries. Design We used the systematic analysis of descriptive epidemiology of 291 diseases and injuries, 1160 sequelae of these diseases and injuries, and 67 risk factors or clusters of risk factors from 1990 to 2010 for 187 countries developed for the Global Burden of Disease 2010 Study to describe the health status of the United States and to compare US health outcomes with those of 34 OECD countries. Years of life lost due to premature mortality (YLLs) were computed by multiplying the number of deaths at each age by a reference life expectancy at that age. Years lived with disability (YLDs) were calculated by multiplying prevalence (based on systematic reviews) by the disability weight (based on population-based surveys) for each sequela; disability in this study refers to any short- or long-term loss of health. Disability-adjusted life-years (DALYs) were estimated as the sum of YLDs and YLLs. Deaths and DALYs related to risk factors were based on systematic reviews and meta-analyses of exposure data and relative risks for risk-outcome pairs. Healthy life expectancy (HALE) was used to summarize overall population health, accounting for both length of life and levels of ill health experienced at different ages. Results US life expectancy for both sexes combined increased from 75.2 years in 1990 to 78.2 years in 2010; during the same period, HALE increased from 65.8 years to 68.1 years. The diseases and injuries with the largest number of YLLs in 2010 were ischemic heart disease, lung cancer, stroke, chronic obstructive pulmonary disease, and road injury. Age-standardized YLL rates increased for Alzheimer disease, drug use disorders, chronic kidney disease, kidney cancer, and falls. The diseases with the largest number of YLDs in 2010 were low back pain, major depressive disorder, other musculoskeletal disorders, neck pain, and anxiety disorders. As the US population has aged, YLDs have comprised a larger share of DALYs than have YLLs. The leading risk factors related to DALYs were dietary risks, tobacco smoking, high body mass index, high blood pressure, high fasting plasma glucose, physical inactivity, and alcohol use. Among 34 OECD countries between 1990 and 2010, the US rank for the age-standardized death rate changed from 18th to 27th, for the age-standardized YLL rate from 23rd to 28th, for the age-standardized YLD rate from 5th to 6th, for life expectancy at birth from 20th to 27th, and for HALE from 14th to 26th. Conclusions and Relevance From 1990 to 2010, the United States made substantial progress in improving health. Life expectancy at birth and HALE increased, all-cause death rates at all ages decreased, and age-specific rates of years lived with disability remained stable. However, morbidity and chronic disability now account for nearly half of the US health burden, and improvements in population health in the United States have not kept pace with advances in population health in other wealthy nations.

Supervised Descent Method and Its Applications to Face Alignment

Abstract: Many computer vision problems (e.g., camera calibration, image alignment, structure from motion) are solved through a nonlinear optimization method. It is generally accepted that 2nd order descent methods are the most robust, fast and reliable approaches for nonlinear optimization of a general smooth function. However, in the context of computer vision, 2nd order descent methods have two main drawbacks: (1) The function might not be analytically differentiable and numerical approximations are impractical. (2) The Hessian might be large and not positive definite. To address these issues, this paper proposes a Supervised Descent Method (SDM) for minimizing a Non-linear Least Squares (NLS) function. During training, the SDM learns a sequence of descent directions that minimizes the mean of NLS functions sampled at different points. In testing, SDM minimizes the NLS objective using the learned descent directions without computing the Jacobian nor the Hessian. We illustrate the benefits of our approach in synthetic and real examples, and show how SDM achieves state-of-the-art performance in the problem of facial feature detection. The code is available at www.humansensing.cs. cmu.edu/intraface.

Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs

Abstract: Most psychiatric disorders are moderately to highly heritable. The degree to which genetic variation is unique to individual disorders or shared across disorders is unclear. To examine shared genetic etiology, we use genome-wide genotype data from the Psychiatric Genomics Consortium (PGC) for cases and controls in schizophrenia, bipolar disorder, major depressive disorder, autism spectrum disorders (ASD) and attention-deficit/hyperactivity disorder (ADHD). We apply univariate and bivariate methods for the estimation of genetic variation within and covariation between disorders. SNPs explained 17-29% of the variance in liability. The genetic correlation calculated using common SNPs was high between schizophrenia and bipolar disorder (0.68 ± 0.04 s.e.), moderate between schizophrenia and major depressive disorder (0.43 ± 0.06 s.e.), bipolar disorder and major depressive disorder (0.47 ± 0.06 s.e.), and ADHD and major depressive disorder (0.32 ± 0.07 s.e.), low between schizophrenia and ASD (0.16 ± 0.06 s.e.) and non-significant for other pairs of disorders as well as between psychiatric disorders and the negative control of Crohn's disease. This empirical evidence of shared genetic etiology for psychiatric disorders can inform nosology and encourages the investigation of common pathophysiologies for related disorders.

The baryon oscillation spectroscopic survey of sdss-iii

Abstract: The Baryon Oscillation Spectroscopic Survey (BOSS) is designed to measure the scale of baryon acoustic oscillations (BAO) in the clustering of matter over a larger volume than the combined efforts of all previous spectroscopic surveys of large-scale structure. BOSS uses 1.5 million luminous galaxies as faint as i = 19.9 over 10,000 deg2 to measure BAO to redshifts z < 0.7. Observations of neutral hydrogen in the Lyα forest in more than 150,000 quasar spectra (g < 22) will constrain BAO over the redshift range 2.15 < z < 3.5. Early results from BOSS include the first detection of the large-scale three-dimensional clustering of the Lyα forest and a strong detection from the Data Release 9 data set of the BAO in the clustering of massive galaxies at an effective redshift z = 0.57. We project that BOSS will yield measurements of the angular diameter distance dA to an accuracy of 1.0% at redshifts z = 0.3 and z = 0.57 and measurements of H(z) to 1.8% and 1.7% at the same redshifts. Forecasts for Lyα forest constraints predict a measurement of an overall dilation factor that scales the highly degenerate DA (z) and H –1(z) parameters to an accuracy of 1.9% at z ~ 2.5 when the survey is complete. Here, we provide an overview of the selection of spectroscopic targets, planning of observations, and analysis of data and data quality of BOSS.

Discrete Signal Processing on Graphs

Abstract: In social settings, individuals interact through webs of relationships. Each individual is a node in a complex network (or graph) of interdependencies and generates data, lots of data. We label the data by its source, or formally stated, we index the data by the nodes of the graph. The resulting signals (data indexed by the nodes) are far removed from time or image signals indexed by well ordered time samples or pixels. DSP, discrete signal processing, provides a comprehensive, elegant, and efficient methodology to describe, represent, transform, analyze, process, or synthesize these well ordered time or image signals. This paper extends to signals on graphs DSP and its basic tenets, including filters, convolution, z-transform, impulse response, spectral representation, Fourier transform, frequency response, and illustrates DSP on graphs by classifying blogs, linear predicting and compressing data from irregularly located weather stations, or predicting behavior of customers of a mobile service provider.

Electrical control of neutral and charged excitons in a monolayer semiconductor

Abstract: Monolayer group-VI transition metal dichalcogenides have recently emerged as semiconducting alternatives to graphene in which the true two-dimensionality is expected to illuminate new semiconducting physics. Here we investigate excitons and trions (their singly charged counterparts), which have thus far been challenging to generate and control in the ultimate two-dimensional limit. Utilizing high-quality monolayer molybdenum diselenide, we report the unambiguous observation and electrostatic tunability of charging effects in positively charged (X+), neutral (Xo) and negatively charged (X−) excitons in field-effect transistors via photoluminescence. The trion charging energy is large (30 meV), enhanced by strong confinement and heavy effective masses, whereas the linewidth is narrow (5 meV) at temperatures <55 K. This is greater spectral contrast than in any known quasi-two-dimensional system. We also find the charging energies for X+ and X− to be nearly identical implying the same effective mass for electrons and holes. Single layers of group-VI transition metal dichalcogenides have emerged as direct bandgap semiconductors in the two-dimensional limit. The authors show that monolayer molybdenum diselenide is an ideal system enabling electrostatic tunability of charging effects in neutral and charged electron-hole pairs, so-called excitons.

Expectation Propagation for approximate Bayesian inference

Abstract: This paper presents a new deterministic approximation technique in Bayesian networks. This method, "Expectation Propagation", unifies two previous techniques: assumed-density filtering, an extension of the Kalman filter, and loopy belief propagation, an extension of belief propagation in Bayesian networks. All three algorithms try to recover an approximate distribution which is close in KL divergence to the true distribution. Loopy belief propagation, because it propagates exact belief states, is useful for a limited class of belief networks, such as those which are purely discrete. Expectation Propagation approximates the belief states by only retaining certain expectations, such as mean and variance, and iterates until these expectations are consistent throughout the network. This makes it applicable to hybrid networks with discrete and continuous nodes. Expectation Propagation also extends belief propagation in the opposite direction - it can propagate richer belief states that incorporate correlations between nodes. Experiments with Gaussian mixture models show Expectation Propagation to be convincingly better than methods with similar computational cost: Laplace's method, variational Bayes, and Monte Carlo. Expectation Propagation also provides an efficient algorithm for training Bayes point machine classifiers.

Optical generation of excitonic valley coherence in monolayer WSe2

Abstract: As a consequence of degeneracies arising from crystal symmetries, it is possible for electron states at band-edges ('valleys') to have additional spin-like quantum numbers. An important question is whether coherent manipulation can be performed on such valley pseudospins, analogous to that implemented using true spin, in the quest for quantum technologies. Here, we show that valley coherence can be generated and detected. Because excitons in a single valley emit circularly polarized photons, linear polarization can only be generated through recombination of an exciton in a coherent superposition of the two valley states. Using monolayer semiconductor WSe2 devices, we first establish the circularly polarized optical selection rules for addressing individual valley excitons and trions. We then demonstrate coherence between valley excitons through the observation of linearly polarized luminescence, whose orientation coincides with that of the linearly polarized excitation, for any given polarization angle. In contrast, the corresponding photoluminescence from trions is not observed to be linearly polarized, consistent with the expectation that the emitted photon polarization is entangled with valley pseudospin. The ability to address coherence, in addition to valley polarization, is a step forward towards achieving quantum manipulation of the valley index necessary for coherent valleytronics.

Derivative-free optimization: a review of algorithms and comparison of software implementations

Abstract: This paper addresses the solution of bound-constrained optimization problems using algorithms that require only the availability of objective function values but no derivative information. We refer to these algorithms as derivative-free algorithms. Fueled by a growing number of applications in science and engineering, the development of derivative-free optimization algorithms has long been studied, and it has found renewed interest in recent time. Along with many derivative-free algorithms, many software implementations have also appeared. The paper presents a review of derivative-free algorithms, followed by a systematic comparison of 22 related implementations using a test set of 502 problems. The test bed includes convex and nonconvex problems, smooth as well as nonsmooth problems. The algorithms were tested under the same conditions and ranked under several criteria, including their ability to find near-global solutions for nonconvex problems, improve a given starting point, and refine a near-optimal solution. A total of 112,448 problem instances were solved. We find that the ability of all these solvers to obtain good solutions diminishes with increasing problem size. For the problems used in this study, TOMLAB/MULTIMIN, TOMLAB/GLCCLUSTER, MCS and TOMLAB/LGO are better, on average, than other derivative-free solvers in terms of solution quality within 2,500 function evaluations. These global solvers outperform local solvers even for convex problems. Finally, TOMLAB/OQNLP, NEWUOA, and TOMLAB/MULTIMIN show superior performance in terms of refining a near-optimal solution.

A Simple, Fast, and Effective Reparameterization of IBM Model 2

Abstract: We present a simple log-linear reparameterization of IBM Model 2 that overcomes problems arising from Model 1’s strong assumptions and Model 2’s overparameterization. Efficient inference, likelihood evaluation, and parameter estimation algorithms are provided. Training the model is consistently ten times faster than Model 4. On three large-scale translation tasks, systems built using our alignment model outperform IBM Model 4. An open-source implementation of the alignment model described in this paper is available from http://github.com/clab/fast align .

Is the Sample Good Enough? Comparing Data from Twitter's Streaming API with Twitter's Firehose

Abstract: Twitter is a social media giant famous for the exchange of short, 140-character messages called "tweets". In the scientific community, the microblogging site is known for openness in sharing its data. It provides a glance into its millions of users and billions of tweets through a "Streaming API" which provides a sample of all tweets matching some parameters preset by the API user. The API service has been used by many researchers, companies, and governmental institutions that want to extract knowledge in accordance with a diverse array of questions pertaining to social media. The essential drawback of the Twitter API is the lack of documentation concerning what and how much data users get. This leads researchers to question whether the sampled data is a valid representation of the overall activity on Twitter. In this work we embark on answering this question by comparing data collected using Twitter's sampled API service with data collected using the full, albeit costly, Firehose stream that includes every single published tweet. We compare both datasets using common statistical metrics as well as metrics that allow us to compare topics, networks, and locations of tweets. The results of our work will help researchers and practitioners understand the implications of using the Streaming API.

Conceptual and Procedural Knowledge : The Case of Mathematics

Abstract: : According to a common folk model, students learn arithmetic by understanding the teacher's explanation of it. This folk model suggests that other, more complicated procedural skills are also acquired by being told. The evidence presented herein suggests that learning-by-being-told is an inaccurate model of the kind of arithmetic learning that actually occurs in classrooms. Rather, arithmetic is learned by induction: the generalization and integration of examples. Contents: 1) Schematic vs. teleological knowledge; 2) Three ways that arithmetic could be learned; 3) The conservative evaluation of the induction hypothesis; 4) A liberal evaluation of the induction hypothesis; 5) Learning by analogy; 6) Learning by being told; 7) Summary; 8) Concluding remarks; 9) Appendix.

The future of crowd work

Abstract: Paid crowd work offers remarkable opportunities for improving productivity, social mobility, and the global economy by engaging a geographically distributed workforce to complete complex tasks on demand and at scale. But it is also possible that crowd work will fail to achieve its potential, focusing on assembly-line piecework. Can we foresee a future crowd workplace in which we would want our children to participate? This paper frames the major challenges that stand in the way of this goal. Drawing on theory from organizational behavior and distributed computing, as well as direct feedback from workers, we outline a framework that will enable crowd work that is complex, collaborative, and sustainable. The framework lays out research challenges in twelve major areas: workflow, task assignment, hierarchy, real-time response, synchronous collaboration, quality control, crowds guiding AIs, AIs guiding crowds, platforms, job design, reputation, and motivation.

Atomically Precise Gold Nanoclusters as New Model Catalysts

Abstract: Many industrial catalysts involve nanoscale metal particles (typically 1–100 nm), and understanding their behavior at the molecular level is a major goal in heterogeneous catalyst research. However, conventional nanocatalysts have a nonuniform particle size distribution, while catalytic activity of nanoparticles is size dependent. This makes it difficult to relate the observed catalytic performance, which represents the average of all particle sizes, to the structure and intrinsic properties of individual catalyst particles. To overcome this obstacle, catalysts with well-defined particle size are highly desirable.In recent years, researchers have made remarkable advances in solution-phase synthesis of atomically precise nanoclusters, notably thiolate-protected gold nanoclusters. Such nanoclusters are composed of a precise number of metal atoms (n) and of ligands (m), denoted as Aun(SR)m, with n ranging up to a few hundred atoms (equivalent size up to 2–3 nm). These protected nanoclusters are well-defined ...

Ligra: a lightweight graph processing framework for shared memory

Abstract: There has been significant recent interest in parallel frameworks for processing graphs due to their applicability in studying social networks, the Web graph, networks in biology, and unstructured meshes in scientific simulation. Due to the desire to process large graphs, these systems have emphasized the ability to run on distributed memory machines. Today, however, a single multicore server can support more than a terabyte of memory, which can fit graphs with tens or even hundreds of billions of edges. Furthermore, for graph algorithms, shared-memory multicores are generally significantly more efficient on a per core, per dollar, and per joule basis than distributed memory systems, and shared-memory algorithms tend to be simpler than their distributed counterparts.In this paper, we present a lightweight graph processing framework that is specific for shared-memory parallel/multicore machines, which makes graph traversal algorithms easy to write. The framework has two very simple routines, one for mapping over edges and one for mapping over vertices. Our routines can be applied to any subset of the vertices, which makes the framework useful for many graph traversal algorithms that operate on subsets of the vertices. Based on recent ideas used in a very fast algorithm for breadth-first search (BFS), our routines automatically adapt to the density of vertex sets. We implement several algorithms in this framework, including BFS, graph radii estimation, graph connectivity, betweenness centrality, PageRank and single-source shortest paths. Our algorithms expressed using this framework are very simple and concise, and perform almost as well as highly optimized code. Furthermore, they get good speedups on a 40-core machine and are significantly more efficient than previously reported results using graph frameworks on machines with many more cores.

Deterministic Diffusion Fiber Tracking Improved by Quantitative Anisotropy

Abstract: Diffusion MRI tractography has emerged as a useful and popular tool for mapping connections between brain regions. In this study, we examined the performance of quantitative anisotropy (QA) in facilitating deterministic fiber tracking. Two phantom studies were conducted. The first phantom study examined the susceptibility of fractional anisotropy (FA), generalized factional anisotropy (GFA), and QA to various partial volume effects. The second phantom study examined the spatial resolution of the FA-aided, GFA-aided, and QA-aided tractographies. An in vivo study was conducted to track the arcuate fasciculus, and two neurosurgeons blind to the acquisition and analysis settings were invited to identify false tracks. The performance of QA in assisting fiber tracking was compared with FA, GFA, and anatomical information from T1-weighted images. Our first phantom study showed that QA is less sensitive to the partial volume effects of crossing fibers and free water, suggesting that it is a robust index. The second phantom study showed that the QA-aided tractography has better resolution than the FA-aided and GFA-aided tractography. Our in vivo study further showed that the QA-aided tractography outperforms the FA-aided, GFA-aided, and anatomy-aided tractographies. In the shell scheme (HARDI), the FA-aided, GFA-aided, and anatomy-aided tractographies have 30.7%, 32.6%, and 24.45% of the false tracks, respectively, while the QA-aided tractography has 16.2%. In the grid scheme (DSI), the FA-aided, GFA-aided, and anatomy-aided tractographies have 12.3%, 9.0%, and 10.93% of the false tracks, respectively, while the QA-aided tractography has 4.43%. The QA-aided deterministic fiber tracking may assist fiber tracking studies and facilitate the advancement of human connectomics.

Mind-Sets Matter: A Meta-Analytic Review of Implicit Theories and Self-Regulation.

Abstract: This review builds on self-control theory (Carver & Scheier, 1998) to develop a theoretical framework for investigating associations of implicit theories with self-regulation. This framework conceptualizes self-regulation in terms of 3 crucial processes: goal setting, goal operating, and goal monitoring. In this meta-analysis, we included articles that reported a quantifiable assessment of implicit theories and at least 1 self-regulatory process or outcome. With a random effects approach used, meta-analytic results (total unique N = 28,217; k = 113) across diverse achievement domains (68% academic) and populations (age range = 5-42; 10 different nationalities; 58% from United States; 44% female) demonstrated that implicit theories predict distinct self-regulatory processes, which, in turn, predict goal achievement. Incremental theories, which, in contrast to entity theories, are characterized by the belief that human attributes are malleable rather than fixed, significantly predicted goal setting (performance goals, r = -.151; learning goals, r = .187), goal operating (helpless-oriented strategies, r = -.238; mastery-oriented strategies, r = .227), and goal monitoring (negative emotions, r = -.233; expectations, r = .157). The effects for goal setting and goal operating were stronger in the presence (vs. absence) of ego threats such as failure feedback. Discussion emphasizes how the present theoretical analysis merges an implicit theory perspective with self-control theory to advance scholarship and unlock major new directions for basic and applied research.

Three-band tight-binding model for monolayers of group-VIB transition metal dichalcogenides

Abstract: We present a three-band tight-binding (TB) model for describing the low-energy physics in monolayers of group-VIB transition metal dichalcogenides $M{X}_{2}$ ($M=\text{Mo}$, W; $X=\text{S}$, Se, Te). As the conduction- and valence-band edges are predominantly contributed by the ${d}_{{z}^{2}}$, ${d}_{xy}$, and ${d}_{{x}^{2}\ensuremath{-}{y}^{2}}$ orbitals of $M$ atoms, the TB model is constructed using these three orbitals based on the symmetries of the monolayers. Parameters of the TB model are fitted from the first-principles energy bands for all $M{X}_{2}$ monolayers. The TB model involving only the nearest-neighbor $M$-$M$ hoppings is sufficient to capture the band-edge properties in the $\ifmmode\pm\else\textpm\fi{}K$ valleys, including the energy dispersions as well as the Berry curvatures. The TB model involving up to the third-nearest-neighbor $M$-$M$ hoppings can well reproduce the energy bands in the entire Brillouin zone. Spin-orbit coupling in valence bands is well accounted for by including the on-site spin-orbit interactions of $M$ atoms. The conduction band also exhibits a small valley-dependent spin splitting which has an overall sign difference between Mo${X}_{2}$ and W${X}_{2}$. We discuss the origins of these corrections to the three-band model. The three-band TB model developed here is efficient to account for low-energy physics in $M{X}_{2}$ monolayers, and its simplicity can be particularly useful in the study of many-body physics and physics of edge states.

Software Engineering for Self-Adaptive Systems : A Second Research Roadmap

Abstract: The goal of this roadmap paper is to summarize the state-of-the-art and identify research challenges when developing, deploying and managing self-adaptive software systems. Instead of dealing with a wide range of topics associated with the field, we focus on four essential topics of self-adaptation: design space for self-adaptive solutions, software engineering processes for self-adaptive systems, from centralized to decentralized control, and practical run-time verification & validation for self-adaptive systems. For each topic, we present an overview, suggest future directions, and focus on selected challenges. This paper complements and extends a previous roadmap on software engineering for self-adaptive systems published in 2009 covering a different set of topics, and reflecting in part on the previous paper. This roadmap is one of the many results of the Dagstuhl Seminar 10431 on Software Engineering for Self-Adaptive Systems, which took place in October 2010.

Improved Part-of-Speech Tagging for Online Conversational Text with Word Clusters

Abstract: We consider the problem of part-of-speech tagging for informal, online conversational text. We systematically evaluate the use of large-scale unsupervised word clustering and new lexical features to improve tagging accuracy. With these features, our system achieves state-of-the-art tagging results on both Twitter and IRC POS tagging tasks; Twitter tagging is improved from 90% to 93% accuracy (more than 3% absolute). Qualitative analysis of these word clusters yields insights about NLP and linguistic phenomena in this genre. Additionally, we contribute the first POS annotation guidelines for such text and release a new dataset of English language tweets annotated using these guidelines. Tagging software, annotation guidelines, and large-scale word clusters are available at: http://www.ark.cs.cmu.edu/TweetNLP This paper describes release 0.3 of the “CMU Twitter Part-of-Speech Tagger” and annotated data. [This paper is forthcoming in Proceedings of NAACL 2013; Atlanta, GA, USA.]

Molecular understanding of sulphuric acid-amine particle nucleation in the atmosphere

Abstract: Nucleation of aerosol particles from trace atmospheric vapours is thought to provide up to half of global cloud condensation nuclei(1). Aerosols can cause a net cooling of climate by scattering sun ...

CHOMP: Covariant Hamiltonian optimization for motion planning

Abstract: In this paper, we present CHOMP (covariant Hamiltonian optimization for motion planning), a method for trajectory optimization invariant to reparametrization. CHOMP uses functional gradient techniques to iteratively improve the quality of an initial trajectory, optimizing a functional that trades off between a smoothness and an obstacle avoidance component. CHOMP can be used to locally optimize feasible trajectories, as well as to solve motion planning queries, converging to low-cost trajectories even when initialized with infeasible ones. It uses Hamiltonian Monte Carlo to alleviate the problem of convergence to high-cost local minima (and for probabilistic completeness), and is capable of respecting hard constraints along the trajectory. We present extensive experiments with CHOMP on manipulation and locomotion tasks, using seven-degree-of-freedom manipulators and a rough-terrain quadruped robot.

Observation of a charged charmoniumlike structure in e+ e- → (D* D*)± π∓ at √s = 4.26 GeV.

Abstract: We study the process e(+)e(-) -> pi(+)pi(-) J/psi at a center-of-mass energy of 4.260 GeV using a 525 pb(-1) data sample collected with the BESIII detector operating at the Beijing Electron Positron Collider. The Born cross section is measured to be (62.9 +/- 1.9 +/- 3.7) pb, consistent with the production of the Y(4260). We observe a structure at around 3.9 GeV/c(2) in the pi(+/-) J/psi mass spectrum, which we refer to as the Z(c)(3900). If interpreted as a new particle, it is unusual in that it carries an electric charge and couples to charmonium. A fit to the pi(+/-) J/psi invariant mass spectrum, neglecting interference, results in a mass of (3899.0 +/- 3.6 +/- 4.9) MeV/c(2) and a width of (46 +/- 10 +/- 20) MeV. Its production ratio is measured to be R = (sigma(e(+)e(-) -> pi(+/-) Z(c)(3900)(-/+) -> pi(+)pi(-) J/psi)/sigma(e(+)e(-) -> pi(+)pi(-) J/psi)) = (21.5 +/- 3.3 +/- 7.5)%. In all measurements the first errors are statistical and the second are systematic.