Perceptual learning, roving and the unsupervised bias

Perceptual learning improves perception through training. Perceptual learning improves with most stimulus types but fails when . certain stimulus types are mixed during training (roving). This result is surprising because classical supervised and unsupervised neural network models can cope easily with roving conditions. What makes humans so inferior compared to these models? As experimental and conceptual work has shown, human perceptual learning is neither supervised nor unsupervised but reward-based learning. Reward-based learning suffers from the so-called unsupervised bias, i.e., to prevent synaptic " drift" , the . average reward has to be exactly estimated. However, this is impossible when two or more stimulus types with different rewards are presented during training (and the reward is estimated by a running average). For this reason, we propose no learning occurs in roving conditions. However, roving hinders perceptual learning only for combinations of similar stimulus types but not for dissimilar ones. In this latter case, we propose that a critic can estimate the reward for each stimulus type separately. One implication of our analysis is that the critic cannot be located in the visual system. © 2011 Elsevier Ltd.

High temperature robust SOI ethanol sensor

In this paper we present a robust SOI-CMOS ethanol sensor based on a tungsten-doped lanthanum iron oxide sensing material. The device shows response to gas, has low power consumption, good uniformity, high temperature stability and can be manufactured at low cost and with integrated circuitry. The platform is a tungsten-based CMOS micro-hotplate that has been shown to be stable for over two thousand hours at a high temperature (600°C) in a form of accelerated life test. The tungsten-doped lanthanum iron oxide was deposited on the micro-hotplate as a slurry with terpineol using a syringe, dried and annealed. Preliminary gas testing was done and the material shows response to ethanol vapour. These results are promising and we believe that this combination of a robust CMOS micro-hotplate and a good sensing material can form the basis for a commercial CMOS gas sensor. © 2011 Published by Elsevier Ltd.

Steady-state modelling of the universal exhaust gas oxygen (UEGO) sensor

The universal exhaust gas oxygen (UEGO) sensor is a well-established device which was developed for the measurement of relative air fuel ratio in internal combustion engines. There is, however, little information available which allows for the prediction of the UEGO's behaviour when exposed to arbitrary gas mixtures, pressures and temperatures. Here we present a steady-state model for the sensor, based on a solution of the Stefan-Maxwell equation, and which includes a momentum balance. The response of the sensor is dominated by a diffusion barrier, which controls the rate of diffusion of gas species between the exhaust and a cavity. Determination of the diffusion barrier characteristics, especially the mean pore size, porosity and tortuosity, is essential for the purposes of modelling, and a measurement technique based on identification of the sensor pressure giving zero temperature sensitivity is shown to be a convenient method of achieving this. The model, suitably calibrated, is shown to make good predictions of sensor behaviour for large variations of pressure, temperature and gas composition. © 2012 IOP Publishing Ltd.

Micro-electro-mechanical resonant tilt sensor

This paper reports a micro-electro-mechanical tilt sensor based on resonant sensing principles. The tilt sensor measures orientation by sensing the component of gravitational acceleration along a specified input axis. Design aspects of the tilt sensor are first introduced and a design trade-off between sensitivity, resolution and robustness is addressed. A prototype sensor is microfabricated in a foundry process. The sensor is characterized to validate predictive analytical and FEA models of performance. The prototype is tested over tilt angles ranging over ±90 degrees and the linearity of the sensor is found to be better than 1.4% over the tilt angle range of ±20°. The noise-limited resolution of the sensor is found to be approximately 0.00026 degrees for an integration time of 0.6 seconds. © 2012 IEEE.

Reinforcement learning using a continuous time actor-critic framework with spiking neurons.

Animals repeat rewarded behaviors, but the physiological basis of reward-based learning has only been partially elucidated. On one hand, experimental evidence shows that the neuromodulator dopamine carries information about rewards and affects synaptic plasticity. On the other hand, the theory of reinforcement learning provides a framework for reward-based learning. Recent models of reward-modulated spike-timing-dependent plasticity have made first steps towards bridging the gap between the two approaches, but faced two problems. First, reinforcement learning is typically formulated in a discrete framework, ill-adapted to the description of natural situations. Second, biologically plausible models of reward-modulated spike-timing-dependent plasticity require precise calculation of the reward prediction error, yet it remains to be shown how this can be computed by neurons. Here we propose a solution to these problems by extending the continuous temporal difference (TD) learning of Doya (2000) to the case of spiking neurons in an actor-critic network operating in continuous time, and with continuous state and action representations. In our model, the critic learns to predict expected future rewards in real time. Its activity, together with actual rewards, conditions the delivery of a neuromodulatory TD signal to itself and to the actor, which is responsible for action choice. In simulations, we show that such an architecture can solve a Morris water-maze-like navigation task, in a number of trials consistent with reported animal performance. We also use our model to solve the acrobot and the cartpole problems, two complex motor control tasks. Our model provides a plausible way of computing reward prediction error in the brain. Moreover, the analytically derived learning rule is consistent with experimental evidence for dopamine-modulated spike-timing-dependent plasticity.

A single FBAR-based temperature and pressure sensors

A novel temperature and pressure sensor based on a single film bulk acoustic resonator (FBAR) is designed. This FBAR support two resonant modes, which response opposite to the change of temperature. By sealed the back cavity of a back-trench membrane type FBAR with silicon wafer, an on-chip single FBAR sensor suitable for measuring temperature and pressure simultaneously is proposed. For unsealed device, the experimental results show that the first resonant mode has a temperature coefficient of frequency (TCF) of 69.5ppm/K, and the TCF of the second mode is -8.1ppm/K. After sealed the back trench, it can be used as a pressure sensor, the pressure coefficient of frequency (PCF) for the two resonant mode is -17.4ppm/kPa and -6.1 ppm/kPa respectively, both of them being more sensitive than other existing pressure sensors. © 2013 Trans Tech Publications Ltd, Switzerland.

Micro-electro-mechanical resonant tilt sensor with 250 nano-radian resolution

This paper reports a high-resolution frequency-output MEMS tilt sensor based on resonant sensing principles. The tilt sensor measures orientation by sensing the component of gravitational acceleration along a specified input axis. A combination of design enhancements enables significantly higher sensitivity for this device as compared to previously reported prototype sensors. The MEMS tilt sensor is calibrated on a manual tilt table over tilt angles ranging over 0-90 degrees with a relatively linear response measured in the range of ±20°(linearity error <2.3%) with a scale factor of approximately 50.06 Hz/degree. The noise-limited resolution of the sensor is found to be approximately 250 nano-radians for an integration time of 0.8 s, which is over an order of magnitude better than previously reported results [1]. © 2013 IEEE.

Thin film monitoring with silicon bulk acoustic resonators

Abstract-This paper reports a single-crystal silicon mass sensor based on a square-plate resonant structure excited in the wine glass bulk acoustic mode at a resonant frequency of 2.065 MHz and an impressive quality factor of 4 million at 12 mtorr pressure. Mass loading on the resonator results in a linear downshift in the resonant frequency of this device, wherein the measured sensitivity is found to be 175 Hz cm2/μg. The silicon resonator is embedded in an oscillator feedback loop, which has a short-term frequency stability of 3 mHz (approximately 1.5 ppb) at an operating pressure of 3.2 mtorr, corresponding to an equivalent mass noise floor of 17 pg/cm2. Possible applications of this device include thin film monitoring and gas sensing, with the potential added benefits of scalability and integration with CMOS technology. © 2008 IEEE.