Neuromorphic hardware databases for exploring structure-function relationships in the brain.

Neuromorphic hardware is the term used to describe full custom-designed integrated circuits, or silicon ''chips'', that are the product of neuromorphic engineering--a methodology for the synthesis of biologically inspired elements and systems, such as individual neurons, retinae, cochleas, oculomotor systems and central pattern generators. We focus on the implementation of neurons and networks of neurons, designed to illuminate structure-function relationships. Neuromorphic hardware can be constructed with either digital or analogue circuitry or with mixed-signal circuitry--a hybrid of the two. Currently, most examples of this type of hardware are constructed using analogue circuits, in complementary metal-oxide-semiconductor technology. The correspondence between these circuits and neurons, or networks of neurons, can exist at a number of levels. At the lowest level, this correspondence is between membrane ion channels and field-effect transistors. At higher levels, the correspondence is between whole conductances and firing behaviour, and filters and amplifiers, devices found in conventional integrated circuit design. Similarly, neuromorphic engineers can choose to design Hodgkin-Huxley model neurons, or reduced models, such as integrate-and-fire neurons. In addition to the choice of level, there is also choice within the design technique itself; for example, resistive and capacitive properties of the neuronal membrane can be constructed with extrinsic devices, or using the intrinsic properties of the materials from which the transistors themselves are composed. So, silicon neurons can be built, with dendritic, somatic and axonal structures, and endowed with ionic, synaptic and morphological properties. Examples of the structure-function relationships already explored using neuromorphic hardware include correlation detection and direction selectivity. Establishing a database for this hardware is valuable for two reasons: first, independently of neuroscientific motivations, the field of neuromorphic engineering would benefit greatly from a resource in which circuit designs could be stored in a form appropriate for reuse and re-fabrication. Analogue designers would benefit particularly from such a database, as there are no equivalents to the algorithmic design methods available to designers of digital circuits. Second, and more importantly for the purpose of this theme issue, is the possibility of a database of silicon neuron designs replicating specific neuronal types and morphologies. In the future, it may be possible to use an automated process to translate morphometric data directly into circuit design compatible formats. The question that needs to be addressed is: what could a neuromorphic hardware database contribute to the wider neuroscientific community that a conventional database could not? One answer is that neuromorphic hardware is expected to provide analogue sensory-motor systems for interfacing the computational power of symbolic, digital systems with the external, analogue environment. It is also expected to contribute to ongoing work in neural-silicon interfaces and prosthetics. Finally, there is a possibility that the use of evolving circuits, using reconfigurable hardware and genetic algorithms, will create an explosion in the number of designs available to the neuroscience community. All this creates the need for a database to be established, and it would be advantageous to set about this while the field is relatively young. This paper outlines a framework for the construction of a neuromorphic hardware database, for use in the biological exploration of structure-function relationships.     

Advancing simulations of biological materials: applications of coarse-grained models on graphics processing unit hardware

The timescales of biological processes, primarily those inherent to the molecular mechanisms of disease, are long (>μs) and involve complex interactions of systems consisting of many atoms (>10⁶). Simulating these systems requires an advanced computational approach, and as such, coarse-grained (CG) models have been developed and highly optimised for accelerator hardware, primarily graphics processing units (GPUs). In this review, I discuss the implementation of CG models for biologically relevant systems, and show how such models can be optimised and perform well on GPU-accelerated hardware. Several examples of GPU implementations of CG models for both molecular dynamics and Monte Carlo simulations on purely GPU and hybrid CPU/GPU architectures are presented. Both the hardware and algorithmic limitations of various models, which depend greatly on the application of interest, are discussed.     

A multi-differential neuromorphic approach to motion detection

This paper presents a multi-differential neuromorphic approach to motion detection. The model is based evidence for a differential operators interpretation of the properties of the cortical motion pathway. We discuss how this strategy, which provides a robust measure of speed for a range of types of image motion using a single computational mechanism, forms a useful framework in which to develop future neuromorphic motion systems. We also discuss both our approaches to developing computational motion models, and constraints in the design strategy for transferring motion models to other domains of early visual processing.     

Fly-like visuomotor responses of a robot using aVLSI motion-sensitive chips

 We explore the use of continuous-time analog very-large-scale-integrated (aVLSI) neuromorphic visual preprocessors together with a robotic platform in generating bio-inspired behaviors. Both the aVLSI motion sensors and the robot behaviors described in this work are inspired by the motion computation in the fly visual system and two different fly behaviors. In most robotic systems, the visual information comes from serially scanned imagers. This restricts the form of computation of the visual image and slows down the input rate to the controller system of the robot, hence increasing the reaction time of the robot. These aVLSI neuromorphic sensors reduce the computational load and power consumption of the robot, thus making it possible to explore continuous-time visuomotor control systems that react in real-time to the environment. The motion sensor provides two outputs: one for the preferred direction and the other for the null direction. These motion outputs are created from the aggregation of six elementary motion detectors that implement a variant of Reichardt's correlation algorithm. The four analog continuous-time outputs from the motion chips go to the control system on the robot which generates a mixture of two behaviors – course stabilization and fixation – from the outputs of these sensors. Since there are only four outputs, the amount of information transmitted to the controller is reduced (as compared to using a CCD sensor), and the reaction time of the robot is greatly decreased. In this work, the robot samples the motion sensors every 3.3 ms during the behavioral experiments. Received: 4 October 1999 / Accepted in revised form: 26 April 2001     

Characterization and Compensation of Network-Level Anomalies in Mixed-Signal Neuromorphic Modeling Platforms

Advancing the size and complexity of neural network models leads to an ever increasing demand for computational resources for their simulation. Neuromorphic devices offer a number of advantages over conventional computing architectures, such as high emulation speed or low power consumption, but this usually comes at the price of reduced configurability and precision. In this article, we investigate the consequences of several such factors that are common to neuromorphic devices, more specifically limited hardware resources, limited parameter configurability and parameter variations due to fixed-pattern noise and trial-to-trial variability. Our final aim is to provide an array of methods for coping with such inevitable distortion mechanisms. As a platform for testing our proposed strategies, we use an executable system specification (ESS) of the BrainScaleS neuromorphic system, which has been designed as a universal emulation back-end for neuroscientific modeling. We address the most essential limitations of this device in detail and study their effects on three prototypical benchmark network models within a well-defined, systematic workflow. For each network model, we start by defining quantifiable functionality measures by which we then assess the effects of typical hardware-specific distortion mechanisms, both in idealized software simulations and on the ESS. For those effects that cause unacceptable deviations from the original network dynamics, we suggest generic compensation mechanisms and demonstrate their effectiveness. Both the suggested workflow and the investigated compensation mechanisms are largely back-end independent and do not require additional hardware configurability beyond the one required to emulate the benchmark networks in the first place. We hereby provide a generic methodological environment for configurable neuromorphic devices that are targeted at emulating large-scale, functional neural networks.     

Critical evaluation of the JDO API for the persistence and portability requirements of complex biological databases

Background  

Complex biological database systems have become key computational tools used daily by scientists and researchers. Many of these systems must be capable of executing on multiple different hardware and software configurations and are also often made available to users via the Internet. We have used the Java Data Object (JDO) persistence technology to develop the database layer of such a system known as the SigPath information management system. SigPath is an example of a complex biological database that needs to store various types of information connected by many relationships.     

Robust Models for Optic Flow Coding in Natural Scenes Inspired by Insect Biology

The extraction of accurate self-motion information from the visual world is a difficult problem that has been solved very efficiently by biological organisms utilizing non-linear processing. Previous bio-inspired models for motion detection based on a correlation mechanism have been dogged by issues that arise from their sensitivity to undesired properties of the image, such as contrast, which vary widely between images. Here we present a model with multiple levels of non-linear dynamic adaptive components based directly on the known or suspected responses of neurons within the visual motion pathway of the fly brain. By testing the model under realistic high-dynamic range conditions we show that the addition of these elements makes the motion detection model robust across a large variety of images, velocities and accelerations. Furthermore the performance of the entire system is more than the incremental improvements offered by the individual components, indicating beneficial non-linear interactions between processing stages. The algorithms underlying the model can be implemented in either digital or analog hardware, including neuromorphic analog VLSI, but defy an analytical solution due to their dynamic non-linear operation. The successful application of this algorithm has applications in the development of miniature autonomous systems in defense and civilian roles, including robotics, miniature unmanned aerial vehicles and collision avoidance sensors.     

Gating in small-animal cardio-thoracic CT

Gating is necessary in cardio-thoracic small-animal imaging because of the physiological motions that are present during scanning. In small-animal computed tomography (CT), gating is mainly performed on a projection base because full scans take much longer than the motion cycle. This paper presents and discusses various gating concepts of small-animal CT, and provides examples of concrete implementation. Since a wide variety of small-animal CT scanner systems exist, scanner systems are discussed with respect to the most suitable gating methods. Furthermore, an overview is given of cardio-thoracic imaging and gating applications. The necessary contrast media are discussed as well as gating limitations. Gating in small-animal imaging requires the acquisition of a gating signal during scanning. This can be done extrinsically (additional hardware, e.g. electrocardiogram) or intrinsically from the projection data itself. The gating signal is used retrospectively during CT reconstruction, or prospectively to trigger parts of the scan. Gating can be performed with respect to the phase or the amplitude of the gating signal, providing different advantages and challenges. Gating methods should be optimized with respect to the diagnostic question, scanner system, animal model, type of narcosis and actual setup. The software-based intrinsic gating approaches increasingly employed give the researcher independence from difficult and expensive hardware changes.     

Neuromorphic adaptive plastic scalable electronics: analog learning systems

Decades of research to build programmable intelligent machines have demonstrated limited utility in complex, real-world environments. Comparing their performance with biological systems, these machines are less efficient by a factor of 1 million1 billion in complex, real-world environments. The Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) program is a multifaceted Defense Advanced Research Projects Agency (DARPA) project that seeks to break the programmable machine paradigm and define a new path for creating useful, intelligent machines. Since real-world systems exhibit infinite combinatorial complexity, electronic neuromorphic machine technology would be preferable in a host of applications, but useful and practical implementations still do not exist. HRL Laboratories LLC has embarked on addressing these challenges, and, in this article, we provide an overview of our project and progress made thus far.     

Modeling Inhibitory Interneurons in Efficient Sensory Coding Models

There is still much unknown regarding the computational role of inhibitory cells in the sensory cortex. While modeling studies could potentially shed light on the critical role played by inhibition in cortical computation, there is a gap between the simplicity of many models of sensory coding and the biological complexity of the inhibitory subpopulation. In particular, many models do not respect that inhibition must be implemented in a separate subpopulation, with those inhibitory interneurons having a diversity of tuning properties and characteristic E/I cell ratios. In this study we demonstrate a computational framework for implementing inhibition in dynamical systems models that better respects these biophysical observations about inhibitory interneurons. The main approach leverages recent work related to decomposing matrices into low-rank and sparse components via convex optimization, and explicitly exploits the fact that models and input statistics often have low-dimensional structure that can be exploited for efficient implementations. While this approach is applicable to a wide range of sensory coding models (including a family of models based on Bayesian inference in a linear generative model), for concreteness we demonstrate the approach on a network implementing sparse coding. We show that the resulting implementation stays faithful to the original coding goals while using inhibitory interneurons that are much more biophysically plausible.     

The linearity of emergent spectro-temporal receptive fields in a model of auditory cortex

The responses of cortical neurons are often characterized by measuring their spectro-temporal receptive fields (STRFs). The STRF of a cell can be thought of as a representation of its stimulus 'preference' but it is also a filter or 'kernel' that represents the best linear prediction of the response of that cell to any stimulus. A range of in vivo STRFs with varying properties have been reported in various species, although none in humans. Using a computational model it has been shown that responses of ensembles of artificial STRFs, derived from limited sets of formative stimuli, preserve information about utterance class and prosody as well as the identity and sex of the speaker in a model speech classification system. In this work we help to put this idea on a biologically plausible footing by developing a simple model thalamo-cortical system built of conductance based neurons and synapses some of which exhibit spike-time-dependent plasticity. We show that the neurons in such a model when exposed to formative stimuli develop STRFs with varying temporal properties exhibiting a range of heterotopic integration. These model neurons also, in common with neurons measured in vivo, exhibit a wide range of non-linearities; this deviation from linearity can be exposed by characterizing the difference between the measured response of each neuron to a stimulus, and the response predicted by the STRF estimated for that neuron. The proposed model, with its simple architecture, learning rule, and modest number of neurons (<1000), is suitable for implementation in neuromorphic analogue VLSI hardware and hence could form the basis of a developmental, real time, neuromorphic sound classification system.     

Application of chaotic dynamics in a recurrent neural network to control: hardware implementation into a novel autonomous roving robot

Originating from a viewpoint that complex/chaotic dynamics would play an important role in biological system including brains, chaotic dynamics introduced in a recurrent neural network was applied to control. The results of computer experiment was successfully implemented into a novel autonomous roving robot, which can only catch rough target information with uncertainty by a few sensors. It was employed to solve practical two-dimensional mazes using adaptive neural dynamics generated by the recurrent neural network in which four prototype simple motions are embedded. Adaptive switching of a system parameter in the neural network results in stationary motion or chaotic motion depending on dynamical situations. The results of hardware implementation and practical experiment using it show that, in given two-dimensional mazes, the robot can successfully avoid obstacles and reach the target. Therefore, we believe that chaotic dynamics has novel potential capability in controlling, and could be utilized to practical engineering application.     

From membranes to systems: self-configuration and self-replication in membrane systems

Membrane systems are purely abstract computational models afar inspired by biological cells, their membranes, and their biochemistry. The inherently parallel nature of membrane systems makes them obviously highly inefficient to execute on a sequential von Neumann computer architecture and in addition, programming a membrane system is often a painstakingly difficult undertaking. The main goal of this paper is to provide some key elements for bringing membrane systems from the abstract model closer to a genuine, novel, and unconventional in silico computer architecture. In particular, we will address the mechanisms of self-configuration and self-replication on a macroscopic level and will discuss some general issues related to genuine hardware realizations on the microscopic level.     

On parameter synthesis by parallel model checking

An important problem in current computational systems biology is to analyze models of biological systems dynamics under parameter uncertainty. This paper presents a novel algorithm for parameter synthesis based on parallel model checking. The algorithm is conceptually universal with respect to the modeling approach employed. We introduce the algorithm, show its scalability, and examine its applicability on several biological models.     

Modeling convergent ON and OFF pathways in the early visual system

For understanding the computation and function of single neurons in sensory systems, one needs to investigate how sensory stimuli are related to a neuron’s response and which biological mechanisms underlie this relationship. Mathematical models of the stimulus–response relationship have proved very useful in approaching these issues in a systematic, quantitative way. A starting point for many such analyses has been provided by phenomenological “linear–nonlinear” (LN) models, which comprise a linear filter followed by a static nonlinear transformation. The linear filter is often associated with the neuron’s receptive field. However, the structure of the receptive field is generally a result of inputs from many presynaptic neurons, which may form parallel signal processing pathways. In the retina, for example, certain ganglion cells receive excitatory inputs from ON-type as well as OFF-type bipolar cells. Recent experiments have shown that the convergence of these pathways leads to intriguing response characteristics that cannot be captured by a single linear filter. One approach to adjust the LN model to the biological circuit structure is to use multiple parallel filters that capture ON and OFF bipolar inputs. Here, we review these new developments in modeling neuronal responses in the early visual system and provide details about one particular technique for obtaining the required sets of parallel filters from experimental data.     

When is self-organization used in biological systems?

Self-organization, or decentralized control, is widespread in biological systems, including cells, organisms, and groups. It is not, however, the universal means of organization. I argue that a biological system will be self-organized when it possesses a large number of subunits, and these subunits lack either the communicational abilities or the computational abilities, or both, that are needed to implement centralized control. Such control requires a well informed and highly intelligent supervisor. I stress that the subunits in a self-organized system do not necessarily have low cognitive abilities. A lack of preadaptations for evolving a system-wide communication network can prevent the evolution of centralized control. Hence, sometimes even systems whose subunits possess high cognitive abilities will be self-organized.     

A biologically plausible acoustic motion detection neural network

In this paper we present an acoustic motion detection system to be used in a small mobile robot. While the first purpose of the system has been to be a reliable computational implementation, cheap enough to be built in hardware, effort has also been taken to construct a biologically plausible solution. The motion detector consists of a neural network composed of motion-direction sensitive neurons with a preferred direction and a preferred region of the azimuth. The system was designed to produce a higher response when stimulated by motion in the preferred direction than in the null direction and that is in fact what the system does, which means that, as desired, the system can detect motion and distinguish its direction.