Background: Particle Swarm Optimization (PSO) is an established method for parameter optimization. It represents a population-based adaptive optimization technique that is influenced by several "strategy parameters". Choosing reasonable parameter values for the PSO is crucial for its convergence behavior, and depends on the optimization task. We present a method for parameter meta-optimization based on PSO and its application to neural network training. The concept of the Optimized Particle Swarm Optimization (OPSO) is to optimize the free parameters of the PSO by having swarms within a swarm. We assessed the performance of the OPSO method on a set of five artificial fitness functions and compared it to the performance of two popular PSO implementations. Results: Our results indicate that PSO performance can be improved if meta-optimized parameter sets are applied. In addition, we could improve optimization speed and quality on the other PSO methods in the majority of our experiments. We applied the OPSO method to neural network training with the aim to build a quantitative model for predicting blood-brain barrier permeation of small organic molecules. On average, training time decreased by a factor of four and two in comparison to the other PSO methods, respectively. By applying the OPSO method, a prediction model showing good correlation with training-, test- and validation data was obtained. Conclusion: Optimizing the free parameters of the PSO method can result in performance gain. The OPSO approach yields parameter combinations improving overall optimization performance. Its conceptual simplicity makes implementing the method a straightforward task.
In this study, we present a highly configurable neuromorphic computing substrate and use it for emulating several types of neural networks. At the heart of this system lies a mixed-signal chip, with analog implementations of neurons and synapses and digital transmission of action potentials. Major advantages of this emulation device, which has been explicitly designed as a universal neural network emulator, are its inherent parallelism and high acceleration factor compared to conventional computers. Its configurability allows the realization of almost arbitrary network topologies and the use of widely varied neuronal and synaptic parameters. Fixed-pattern noise inherent to analog circuitry is reduced by calibration routines. An integrated development environment allows neuroscientists to operate the device without any prior knowledge of neuromorphic circuit design. As a showcase for the capabilities of the system, we describe the successful emulation of six different neural networks which cover a broad spectrum of both structure and functionality.
To internally reflect the sensory environment, animals create neural maps encoding the external stimulus space. From that primary neural code relevant information has to be extracted for accurate navigation. We analyzed how different odor features such as hedonic valence and intensity are functionally integrated in the lateral horn (LH) of the vinegar fly, Drosophila melanogaster. We characterized an olfactory-processing pathway, comprised of inhibitory projection neurons (iPNs) that target the LH exclusively, at morphological, functional and behavioral levels. We demonstrate that iPNs are subdivided into two morphological groups encoding positive hedonic valence or intensity information and conveying these features into separate domains in the LH. Silencing iPNs severely diminished flies' attraction behavior. Moreover, functional imaging disclosed a LH region tuned to repulsive odors comprised exclusively of third-order neurons. We provide evidence for a feature-based map in the LH, and elucidate its role as the center for integrating behaviorally relevant olfactory information.
27 January 2014 in Proceedings of the National Academy of Sciences
Significance: One primary goal of computational neuroscience is to uncover fundamental principles of computations that are performed by the brain. In our work, we took direct inspiration from biology for a technical application of brain-like processing. We make use of neuromorphic hardware—electronic versions of neurons and synapses on a microchip—to implement a neural network inspired by the sensory processing architecture of the nervous system of insects. We demonstrate that this neuromorphic network achieves classification of generic multidimensional data—a widespread problem with many technical applications. Our work provides a proof of concept for using analog electronic microcircuits mimicking neurons to perform real-world computing tasks, and it describes the benefits and challenges of the neuromorphic approach.
1 November 2016 in Sensors and Actuators B: Chemical
Estimating the distance of a gas source is important in many applications of chemical sensing, like e.g. environmental monitoring, or chemically-guided robot navigation. If an estimation of the gas concentration at the source is available, source proximity can be estimated from the time-averaged gas concentration at the sensing site. However, in turbulent environments, where fast concentration fluctuations dominate, comparably long measurements are required to obtain a reliable estimate. A lesser known feature that can be exploited for distance estimation in a turbulent environment lies in the relationship between source proximity and the temporal variance of the local gas concentration - the farther the source, the more intermittent are gas encounters. However, exploiting this feature requires measurement of changes in gas concentration on a comparably fast time scale, that have up to now only been achieved using photo-ionisation detectors. Here, we demonstrate that by appropriate signal processing, off-the-shelf metal-oxide sensors are capable of extracting rapidly fluctuating features of gas plumes that strongly correlate with source distance. We show that with a straightforward analysis method it is possible to decode events of large, consistent changes in the measured signal, so-called 'bouts'. The frequency of these bouts predicts the distance of a gas source in wind-tunnel experiments with good accuracy. In addition, we found that the variance of bout counts indicates cross-wind offset to the centreline of the gas plume. Our results offer an alternative approach to estimating gas source proximity that is largely independent of gas concentration, using off-the-shelf metal-oxide sensors. The analysis method we employ demands very few computational resources and is suitable for low-power microcontrollers.Comment: 41 pages, 20 figure