Design and testing of a 250 kW medium-speed Brushless DFIG

This paper presents the design and testing of a 250 kW medium-speed Brushless Doubly-Fed Induction Generator (Brushless DFIG), and its associated power electronics and control systems. The experimental tests confirm the design, and show the system's steady-state and dynamic performance. The medium-speed Brushless DFIG in combination with a simplified two-stage gearbox promises a low-cost low-maintenance and reliable drive train for wind turbine applications.

Machine learning for predictive condensed-phase simulation

We show how machine learning techniques based on Bayesian inference can be used to reach new levels of realism in the computer simulation of molecular materials, focusing here on water. We train our machine-learning algorithm using accurate, correlated quantum chemistry, and predict energies and forces in molecular aggregates ranging from clusters to solid and liquid phases. The widely used electronic-structure methods based on density-functional theory (DFT) give poor accuracy for molecular materials like water, and we show how our techniques can be used to generate systematically improvable corrections to DFT. The resulting corrected DFT scheme gives remarkably accurate predictions for the relative energies of small water clusters and of different ice structures, and greatly improves the description of the structure and dynamics of liquid water.

N-best error simulation for training spoken dialogue systems

A recent trend in spoken dialogue research is the use of reinforcement learning to train dialogue systems in a simulated environment. Past researchers have shown that the types of errors that are simulated can have a significant effect on simulated dialogue performance. Since modern systems typically receive an N-best list of possible user utterances, it is important to be able to simulate a full N-best list of hypotheses. This paper presents a new method for simulating such errors based on logistic regression, as well as a new method for simulating the structure of N-best lists of semantics and their probabilities, based on the Dirichlet distribution. Off-line evaluations show that the new Dirichlet model results in a much closer match to the receiver operating characteristics (ROC) of the live data. Experiments also show that the logistic model gives confusions that are closer to the type of confusions observed in live situations. The hope is that these new error models will be able to improve the resulting performance of trained dialogue systems. © 2012 IEEE.

Policy optimisation of POMDP-based dialogue systems without state space compression

The partially observable Markov decision process (POMDP) has been proposed as a dialogue model that enables automatic improvement of the dialogue policy and robustness to speech understanding errors. It requires, however, a large number of dialogues to train the dialogue policy. Gaussian processes (GP) have recently been applied to POMDP dialogue management optimisation showing an ability to substantially increase the speed of learning. Here, we investigate this further using the Bayesian Update of Dialogue State dialogue manager. We show that it is possible to apply Gaussian processes directly to the belief state, removing the need for a parametric policy representation. In addition, the resulting policy learns significantly faster while maintaining operational performance. © 2012 IEEE.

Code-specific policy gradient rules for spiking neurons

Although it is widely believed that reinforcement learning is a suitable tool for describing behavioral learning, the mechanisms by which it can be implemented in networks of spiking neurons are not fully understood. Here, we show that different learning rules emerge from a policy gradient approach depending on which features of the spike trains are assumed to influence the reward signals, i.e., depending on which neural code is in effect. We use the framework of Williams (1992) to derive learning rules for arbitrary neural codes. For illustration, we present policy-gradient rules for three different example codes - a spike count code, a spike timing code and the most general "full spike train" code - and test them on simple model problems. In addition to classical synaptic learning, we derive learning rules for intrinsic parameters that control the excitability of the neuron. The spike count learning rule has structural similarities with established Bienenstock-Cooper-Munro rules. If the distribution of the relevant spike train features belongs to the natural exponential family, the learning rules have a characteristic shape that raises interesting prediction problems.

Hybrid and passive mode-locking of a monolithic two-section MQW InGaN/GaN laser diode

We report the first hybrid mode-locking of a monolithic two-section multiple quantum well InGaN based laser diode. This device, with a length of 1.5 mm, has a 50-μm-long absorber section located at the back facet and generates a continuous stable 28.6 GHz pulse train with an average output power of 9.4 mW at an emission wavelength of 422 nm. Under hybrid mode-locking, the pulse width reduces to 4 ps, the peak power increases to 72 mW, and the microwave linewidth reduces by 13 dB to <500 kHz. We also observe the passive mode-locking with pulse width and peak power of 8 ps and 37 mW, respectively. © 1989-2012 IEEE.

High power, low-jitter encoded picosecond pulse generation using an rf-locked self-q-switched multicontact GaAs/GaAlAs diode laser

The RF locking of a self-Q-switching diode laser is shown to reduce the jitter of a 2.48 GHz train of 1 W peak power picosecond pulses to less than 300 fs. By using direct modulation of the loss in the Q-switched laser, direct encoding of data has been achieved at rates in excess of 2 Gbit/s.

Characteristics of the spectral dynamics of picosecond modulated AlGaAs/GaAs injection lasers

Breakdown of the optical spectrum of a train of picosecond pulses into components with a distance which exceeds kT (200 cm-1 at λ = 955 nm and T = 300 K) is discovered for the first time in an injection laser. The effect may be caused by combined interaction between photons and phonons, with collective excitations in the degraded electron-hole GaAs plasma, and with the stream of drifting carriers in the active medium of the laser.

Design and testing of a medium-speed 250 kW Brushless DFIG

This paper presents the design and testing of a 250 kW medium-speed Brushless Doubly-Fed Generator (Brushless DFIG), and its associated power electronics and control systems. The experimental tests confirm the design, and show the system's steady-state and dynamic performance. The medium-speed Brushless DFIG in combination with a simplified twostage gearbox promises a low-cost low-maintenance and reliable drive train for wind turbine applications.

Picosecond pulse generation in narrow stripe mode-locked quantum dot master oscillator power amplifier

This paper reports a monolithically integrated mode-locked narrow stripe QD MOPA operating at 1300nm generating a stable 20GHz pulse train with an average power of 46.4mW and a pulse duration of 8.3ps. © Optical Society of America.

Gaussian processes for POMDP-based dialogue manager optimization

A partially observable Markov decision process (POMDP) has been proposed as a dialog model that enables automatic optimization of the dialog policy and provides robustness to speech understanding errors. Various approximations allow such a model to be used for building real-world dialog systems. However, they require a large number of dialogs to train the dialog policy and hence they typically rely on the availability of a user simulator. They also require significant designer effort to hand-craft the policy representation. We investigate the use of Gaussian processes (GPs) in policy modeling to overcome these problems. We show that GP policy optimization can be implemented for a real world POMDP dialog manager, and in particular: 1) we examine different formulations of a GP policy to minimize variability in the learning process; 2) we find that the use of GP increases the learning rate by an order of magnitude thereby allowing learning by direct interaction with human users; and 3) we demonstrate that designer effort can be substantially reduced by basing the policy directly on the full belief space thereby avoiding ad hoc feature space modeling. Overall, the GP approach represents an important step forward towards fully automatic dialog policy optimization in real world systems. © 2013 IEEE.

270GHz, 580fs Optical Pulse Generation from a single-section quantum-dash fabry-pérot laser using frequency multiplication

Pulse generation from a mode-locked single-section 1.55μm quantum-dash FP laser is demonstrated under continuous-wave operation. A 270GHz, 580fs pulse train is achieved by applying frequency multiplication using fiber dispersion. ©2009 Optical Society of America.

Accuracy and transferability of GAP models for tungsten

We introduce interatomic potentials for tungsten in the bcc crystal phase and its defects within the Gaussian Approximation Potential (GAP) framework, fitted to a database of first principles density functional theory (DFT) calculations. We investigate the performance of a sequence of models based on databases of increasing coverage in configuration space and showcase our strategy of choosing representative small unit cells to train models that predict properties only observable using thousands of atoms. The most comprehensive model is then used to calculate properties of the screw dislocation, including its structure, the Peierls barrier and the energetics of the vacancy-dislocation interaction. All software and raw data are available at www.libatoms.org.

Exploiting passive dynamics for robot throwing task

Throwing is a complex and highly dynamic task. Humans usually exploit passive dynamics of their limbs to optimize their movement and muscle activation. In order to approach human throwing, we developed a double pendulum robotic platform. To introduce passivity into the actuated joints, clutches were included in the drive train. In this paper, we demonstrate the advantage of exploiting passive dynamics in reducing the mechanical work. However, engaging and disengaging the clutches are done in discrete fashions. Therefore, we propose an optimization approach which can deal with such discontinuities. It is shown that properly engaging/disengaging the clutches can reduce the mechanical work of a throwing task. The result is compared to the solution of fully actuated double pendulum, both in simulation and experiment. © 2012 IEEE.