Locally weighted learning model predictivek control for elastic joint robots

This paper proposes an efficient and online learning control system that uses the successful Model Predictive Control (MPC) method in a model based locally weighted learning framework. The new approach named Locally Weighted Learning Model Predictive Control (LWL-MPC) has been proposed as a solution to learn to control complex and nonlinear Elastic Joint Robots (EJR). Elastic Joint Robots are generally difficult to learn to control due to their elastic properties preventing standard model learning techniques from being used, such as learning computed torque control. This paper demonstrates the capability of LWL-MPC to perform online and incremental learning while controlling the joint positions of a real three Degree of Freedom (DoF) EJR. An experiment on a real EJR is presented and LWL-MPC is shown to successfully learn to control the system to follow two different figure of eight trajectories.

Locally weighted learning model predictive control for nonlinear and time varying dynamics

This paper proposes an online learning control system that uses the strategy of Model Predictive Control (MPC) in a model based locally weighted learning framework. The new approach, named Locally Weighted Learning Model Predictive Control (LWL-MPC), is proposed as a solution to learn to control robotic systems with nonlinear and time varying dynamics. This paper demonstrates the capability of LWL-MPC to perform online learning while controlling the joint trajectories of a low cost, three degree of freedom elastic joint robot. The learning performance is investigated in both an initial learning phase, and when the system dynamics change due to a heavy object added to the tool point. The experiment on the real elastic joint robot is presented and LWL-MPC is shown to successfully learn to control the system with and without the object. The results highlight the capability of the learning control system to accommodate the lack of mechanical consistency and linearity in a low cost robot arm.

Locally weighted learning methods for non-rigid robot control

This thesis develops a novel approach to robot control that learns to account for a robot's dynamic complexities while executing various control tasks using inspiration from biological sensorimotor control and machine learning. A robot that can learn its own control system can account for complex situations and adapt to changes in control conditions to maximise its performance and reliability in the real world. This research has developed two novel learning methods, with the aim of solving issues with learning control of non-rigid robots that incorporate additional dynamic complexities. The new learning control system was evaluated on a real three degree-of-freedom elastic joint robot arm with a number of experiments: initially validating the learning method and testing its ability to generalise to new tasks, then evaluating the system during a learning control task requiring continuous online model adaptation.

Adding a receding horizon to Locally Weighted Regression for learning robot control

There have been notable advances in learning to control complex robotic systems using methods such as Locally Weighted Regression (LWR). In this paper we explore some potential limits of LWR for robotic applications, particularly investigating its application to systems with a long horizon of temporal dependence. We define the horizon of temporal dependence as the delay from a control input to a desired change in output. LWR alone cannot be used in a temporally dependent system to find meaningful control values from only the current state variables and output, as the relationship between the input and the current state is under-constrained. By introducing a receding horizon of the future output states of the system, we show that sufficient constraint is applied to learn good solutions through LWR. The new method, Receding Horizon Locally Weighted Regression (RH-LWR), is demonstrated through one-shot learning on a real Series Elastic Actuator controlling a pendulum.

Learning multidimensional joint control of a robot using receding horizon locally weighted regression

In this paper we explore the ability of a recent model-based learning technique Receding Horizon Locally Weighted Regression (RH-LWR) useful for learning temporally dependent systems. In particular this paper investigates the application of RH-LWR to learn control of Multiple-input Multiple-output robot systems. RH-LWR is demonstrated through learning joint velocity and position control of a three Degree of Freedom (DoF) rigid body robot.

Fast Pose Estimation with Parameter Sensitive Hashing

Example-based methods are effective for parameter estimation problems when the underlying system is simple or the dimensionality of the input is low. For complex and high-dimensional problems such as pose estimation, the number of required examples and the computational complexity rapidly becme prohibitively high. We introduce a new algorithm that learns a set of hashing functions that efficiently index examples relevant to a particular estimation task. Our algorithm extends a recently developed method for locality-sensitive hashing, which finds approximate neighbors in time sublinear in the number of examples. This method depends critically on the choice of hash functions; we show how to find the set of hash functions that are optimally relevant to a particular estimation problem. Experiments demonstrate that the resulting algorithm, which we call Parameter-Sensitive Hashing, can rapidly and accurately estimate the articulated pose of human figures from a large database of example images.

Active Learning with Statistical Models

For many types of learners one can compute the statistically 'optimal' way to select data. We review how these techniques have been used with feedforward neural networks. We then show how the same principles may be used to select data for two alternative, statistically-based learning architectures: mixtures of Gaussians and locally weighted regression. While the techniques for neural networks are expensive and approximate, the techniques for mixtures of Gaussians and locally weighted regression are both efficient and accurate.

Prediction Of Hydrological Models’ Uncertainty By A Committee Of Machine Learning-Models

This study presents an approach to combine uncertainties of the hydrological model outputs predicted from a number of machine learning models. The machine learning based uncertainty prediction approach is very useful for estimation of hydrological models' uncertainty in particular hydro-metrological situation in real-time application [1]. In this approach the hydrological model realizations from Monte Carlo simulations are used to build different machine learning uncertainty models to predict uncertainty (quantiles of pdf) of the a deterministic output from hydrological model . Uncertainty models are trained using antecedent precipitation and streamflows as inputs. The trained models are then employed to predict the model output uncertainty which is specific for the new input data. We used three machine learning models namely artificial neural networks, model tree, locally weighted regression to predict output uncertainties. These three models produce similar verification results, which can be improved by merging their outputs dynamically. We propose an approach to form a committee of the three models to combine their outputs. The approach is applied to estimate uncertainty of streamflows simulation from a conceptual hydrological model in the Brue catchment in UK and the Bagmati catchment in Nepal. The verification results show that merged output is better than an individual model output. [1] D. L. Shrestha, N. Kayastha, and D. P. Solomatine, and R. Price. Encapsulation of parameteric uncertainty statistics by various predictive machine learning models: MLUE method, Journal of Hydroinformatic, in press, 2013.

Aktionenlernen mit Selbstorganisierenden Karten und Reinforcement Learning

Die vorliegende Arbeit beschäftigt sich mit der Entwicklung eines Funktionsapproximators und dessen Verwendung in Verfahren zum Lernen von diskreten und kontinuierlichen Aktionen: 1. Ein allgemeiner Funktionsapproximator – Locally Weighted Interpolating Growing Neural Gas (LWIGNG) – wird auf Basis eines Wachsenden Neuralen Gases (GNG) entwickelt. Die topologische Nachbarschaft in der Neuronenstruktur wird verwendet, um zwischen benachbarten Neuronen zu interpolieren und durch lokale Gewichtung die Approximation zu berechnen. Die Leistungsfähigkeit des Ansatzes, insbesondere in Hinsicht auf sich verändernde Zielfunktionen und sich verändernde Eingabeverteilungen, wird in verschiedenen Experimenten unter Beweis gestellt. 2. Zum Lernen diskreter Aktionen wird das LWIGNG-Verfahren mit Q-Learning zur Q-LWIGNG-Methode verbunden. Dafür muss der zugrunde liegende GNG-Algorithmus abgeändert werden, da die Eingabedaten beim Aktionenlernen eine bestimmte Reihenfolge haben. Q-LWIGNG erzielt sehr gute Ergebnisse beim Stabbalance- und beim Mountain-Car-Problem und gute Ergebnisse beim Acrobot-Problem. 3. Zum Lernen kontinuierlicher Aktionen wird ein REINFORCE-Algorithmus mit LWIGNG zur ReinforceGNG-Methode verbunden. Dabei wird eine Actor-Critic-Architektur eingesetzt, um aus zeitverzögerten Belohnungen zu lernen. LWIGNG approximiert sowohl die Zustands-Wertefunktion als auch die Politik, die in Form von situationsabhängigen Parametern einer Normalverteilung repräsentiert wird. ReinforceGNG wird erfolgreich zum Lernen von Bewegungen für einen simulierten 2-rädrigen Roboter eingesetzt, der einen rollenden Ball unter bestimmten Bedingungen abfangen soll.

Building a soil spectral library for Tajikistan comparing local and global modeling approaches

Over the last ~20 years, soil spectral libraries storing near-infrared reflectance (NIR) spectra from diverse soil samples have been built for many places, since almost 10 years also for Tajikistan. Many calibration approaches have been reported and used for prediction from large and heterogeneous libraries, but most are hampered by the high diversity of the soils, where the mineral background is heavily influencing spectral features. In such cases, local learning strategies have the advantage of building locally adapted calibrations, which can deal better with nonlinearities. Therefore, it was our major aim to identify the most efficient approach to develop an accurate and stable locally weigthed calibration model using a spectral library compiled over the past years. Keywords: Tajikistan, Near-Infrared spectroscopy (NIRS), soil organic carbon, locally weighted regression, regional and local spectral library.

Estimation of the abundance of yellowfin tuna, Thunnus albacares, by age groups and regions within the Eastern Pacific Ocean

ENGLISH: Monthly estimates of the abundance of yellowfin tuna by age groups and regions within the eastern Pacific Ocean during 1970-1988 are made, using purse-seine catch rates, length-frequency samples, and results from cohort analysis. The numbers of individuals caught of each age group in each logged purse-seine set are estimated, using the tonnage from that set and length-frequency distribution from the "nearest" length-frequency sample(s). Nearest refers to the closest length frequency sample(s) to the purse-seine set in time, distance, and set type (dolphin associated, floating object associated, skipjack associated, none of these, and some combinations). Catch rates are initially calculated as the estimated number of individuals of the age group caught per hour of searching. Then, to remove the effects of set type and vessel speed, they are standardized, using separate weiznted generalized linear models for each age group. The standardized catch rates at the center of each 2.5 0 quadrangle-month are estimated, using locally-weighted least-squares regressions on latitude, longitude and date, and then combined into larger regions. Catch rates within these regions are converted to numbers of yellowfin, using the mean age composition from cohort analysis. The variances of the abundance estimates within regions are large for 0-, 1-, and 5-year-olds, but small for 1.5- to 4-year-olds, except during periods of low fishing activity. Mean annual catch rate estimates for the entire eastern Pacific Ocean are significantly positively correlated with mean abundance estimates from cohort analysis for age groups ranging from 1.5 to 4 years old. Catch-rate indices of abundance by age are expected to be useful in conjunction with data on reproductive biology to estimate total egg production within regions. The estimates may also be useful in understanding geographic and temporal variations in age-specific availability to purse seiners, as well as age-specific movements. SPANISH: Se calculan estimaciones mensuales de la abundancia del atún aleta amarilla por grupos de edad y regiones en el Océano Pacífico oriental durante 1970-1988, usando tasas de captura cerquera, muestras de frecuencia de talla, y los resultados del análisis de cohortes. Se estima el número de individuos capturados de cada grupo de edad en cada lance cerquero registrado, usando el tonelaje del lance en cuestión y la distribución de frecuencia de talla de la(s) muestra(s) de frecuencia de talla "más cercana/s)," "Más cercana" significa la(s) muestra(s) de frecuencia de talla más parecida(s) al lance cerquero en cuanto a fecha, distancia, y tipo de lance (asociado con delfines, con objeto flotante, con barrilete, con ninguno de éstos, y algunas combinaciones). Se calculan inicialmente las tasas de captura como el número estimado de individuos del grupo de edad capturado por hora de búsqueda. A continuación, para eliminar los efectos del tipo de lance y la velocidad del barco, se estandardizan dichas tasas, usando un modelo lineal generalizado ponderado, para cada grupo por separado. Se estima la tasa de captura estandardizada al centro de cada cuadrángulo de 2.5°-mes, usando regresiones de mínimos cuadrados ponderados localmente por latitud, longitud, y fecha, y entonces combinándolas en regiones mayores. Se convierten las tasas de captura dentro de estas regiones en números de aletas amarillas individuales, usando el número promedio por edad proveniente del análisis de cohortes. Las varianzas de las estimaciones de la abundancia dentro de las regiones son grandes para los peces de O, 1, Y5 años de edad, pero pequeñas para aquellos de entre 1.5 Y4 años de edad, excepto durante períodos de poca actividad pesquera. Las estimaciones de la tasa de captura media anual para todo el Océano Pacífico oriental están correlacionadas positivamente de forma significativa con las estimaciones de la abundancia media del análisis de las cohortes para los grupos de edad de entre 1.5 y 4 años. Se espera que los índices de abundancia por edad basados en las tasas de captura sean útiles, en conjunto con datos de la biología reproductiva, para estimar la producción total de huevos por regiones. Las estimaciones podrían asimismo ser útiles para la comprensión de las variaciones geográficas y temporales de la disponibilidad específica por edad a los barcos cerqueros, y también las migraciones específicas por edad. (PDF contains 35 pages.)

Minimizing Statistical Bias with Queries

I describe an exploration criterion that attempts to minimize the error of a learner by minimizing its estimated squared bias. I describe experiments with locally-weighted regression on two simple kinematics problems, and observe that this "bias-only" approach outperforms the more common "variance-only" exploration approach, even in the presence of noise.

Support vector machines for regression and applications to software quality prediction

Software metrics are the key tool in software quality management. In this paper, we propose to use support vector machines for regression applied to software metrics to predict software quality. In experiments we compare this method with other regression techniques such as Multivariate Linear Regression, Conjunctive Rule and Locally Weighted Regression. Results on benchmark dataset MIS, using mean absolute error, and correlation coefficient as regression performance measures, indicate that support vector machines regression is a promising technique for software quality prediction. In addition, our investigation of PCA based metrics extraction shows that using the first few Principal Components (PC) we can still get relatively good performance.

Robust fuzzy Gustafson-Kessel clustering for nonlinear system identification

This paper deals with Takagi-Sugeno (TS) fuzzy model identification of nonlinear systems using fuzzy clustering. In particular, an extended fuzzy Gustafson-Kessel (EGK) clustering algorithm, using robust competitive agglomeration (RCA), is developed for automatically constructing a TS fuzzy model from system input-output data. The EGK algorithm can automatically determine the 'optimal' number of clusters from the training data set. It is shown that the EGK approach is relatively insensitive to initialization and is less susceptible to local minima, a benefit derived from its agglomerate property. This issue is often overlooked in the current literature on nonlinear identification using conventional fuzzy clustering. Furthermore, the robust statistical concepts underlying the EGK algorithm help to alleviate the difficulty of cluster identification in the construction of a TS fuzzy model from noisy training data. A new hybrid identification strategy is then formulated, which combines the EGK algorithm with a locally weighted, least-squares method for the estimation of local sub-model parameters. The efficacy of this new approach is demonstrated through function approximation examples and also by application to the identification of an automatic voltage regulation (AVR) loop for a simulated 3 kVA laboratory micro-machine system.

Centennial-scale climate change in Ireland during the Holocene

We examine mid- to late Holocene centennial-scale climate variability in Ireland using proxy data from peatlands, lakes and a speleothem. A high degree of between-record variability is apparent in the proxy data and significant chronological uncertainties are present. However, tephra layers provide a robust tool for correlation and improve the chronological precision of the records. Although we can find no statistically significant coherence in the dataset as a whole, a selection of high-quality peatland water table reconstructions co-vary more than would be expected by chance alone. A locally weighted regression model with bootstrapping can be used to construct a ‘best-estimate’ palaeoclimatic reconstruction from these datasets. Visual comparison and cross-wavelet analysis of peatland water table compilations from Ireland and Northern Britain show that there are some periods of coherence between these records. Some terrestrial palaeoclimatic changes in Ireland appear to coincide with changes in the North Atlantic thermohaline circulation and solar activity. However, these relationships are inconsistent and may be obscured by chronological uncertainties. We conclude by suggesting an agenda for future Holocene climate research in Ireland. ©2013 Elsevier B.V. All rights reserved.