Can one develop an abstract description of the dynamics of pattern generators that provides quantitative insight into their operation? We explored this question by examining the dynamics of a model central pattern generator that was created using an evolutionary algorithm. We propose an abstract description based on the concept of a dynamical module, a set of neurons that simultaneously make their transitions from one quasistable state to another while the synaptic inputs that they receive remain essentially constant, thus temporarily reducing the dimensionality of the circuit dynamics. Using the mathematical tools of dynamical systems theory, we describe a method for identifying dynamical modules, and demonstrate that this concept can be used to quantitatively characterize constraints on neural architecture, account for phase durations, and predict the effects of parameter changes. Moreover, this abstract description reveals coordinated parameter changes that leave the overall circuit...

ABSTRACT Recent experiments indicate that the dark-adapted vertebrate visual system can count photons with a reliability limited by dark noise in the rod photoreceptors themselves. This suggests that subsequent layers of the retina, responsible for signal processing, add little if any excess noise and extract all the available information. Given the signal and noise characteristics of the photoreceptors, what is the structure of such an optimal processor?We show that optimal estimates of time-varying light intensity can be accomplished by a two-stage filter, and we suggest that the first stage should be identified with the filtering which occurs at the first anatomical stage in retinal signal processing, signal transfer from the rod photoreceptor to the bipolar cell. This leads to parameter-free predictions of the bipolar cell response, which are in excellent agreement with experiments comparing rod and bipolar cell dynamics in the same retina. As far as we know this is the first case in which the computationally significant dynamics of a neuron could be predicted rather than modeled.

Abstract—Perception, prediction and generation of sequences is a fundamental aspect of human behavior and depends on the ability to detect serial order. This paper presents a plausible model of sequential memory at the neurological level based on the theory of cell assemblies. The basic idea is that sequences in the brain are represented by cell assemblies. Each item of the sequence and the sequential association between the items are represented by cell assemblies. Simulation results show that the model is capable of recognizing and discriminating multiple sequences stored in memory. The cell assemblies that represent the sequential association between two items are activated if these items occur in the input in the correct order. These sequence detecting cell assemblies form the basis of this model. A simulation presenting 100 stored sequences and 100 not stored recognizes perfectly 90 % of the time with no false positives.

Sequence learning is arguably the most prevalent form of human and animal learning. Sequences play a pivotal role in classical studies of instrumental conditioning,[1] in human skill learning, and in human high-level problem solving and reasoning. So, it's logical that sequence learning is an important component of learning in many task domains of intelligent systems: inference, planning, reasoning, robotics, natural language processing...

This paper studies the behavior of a large body of neurons in the continuum limit. A mathematical characterization of such systems is obtained by approximating the inverse input-output nonlinearity of a cell (or an assembly of cells) by three adjustable linearized sections. The associative spatio-temporal patterns for storage in the neural system are obtained by using approaches analogous to solving space-time field equations in physics. A noise-reducing equation is also derived from this neural model. In addition, conditions that make a noisy pattern retrievable are identified. Based on these analyses, a visual cortex model is proposed and an exact characterization of the patterns that are storable in this cortex is obtained. Furthermore, we show that this model achieves pattern association that is invariant to scaling, translation, rotation and mirror-reflection. 1 Please contact hjchang@pine.ece.utexas.edu or ghosh@pine.ece.utexas.edu for any questions or comments regarding this p...

An accumulation of anatomical, behavioral, and electrophysiological evidence allows us to identify the neuronal circuitry that is involved with vibrissa-mediated sensation and the control of rhythmic vibrissa movement. Anatomical evidence points to a multiplicity of closed sensorimotor loops, while electrophysiological data delineate the flow of electrical signals in these pathways. These loops process sensory input from the vibrissae and send projections to direct vibrissa movement, starting at the level of the hindbrain and proceeding toward loops that involve multiple structures in the forebrain. The nature of the vibrissa-related electrical signals in behaving animals has been studied extensively at the level of neocortical loops. Two types of spike signal are observed that serve as a reference of vibrissa motion: a fast signal that correlates with the relative phase of the vibrissae within a whisk cycle and a slow signal that correlates with the amplitude, and possibly the set-point, of the vibrissae during a whisk. Both signals are observed in vibrissa primary sensory (S1) cortex, and in some cases they are sufficiently robust to allow vibrissa position to be accurately estimated from the spike train of a single neuron. Unlike the case for S1 cortex, only the slow signal has been observed in vibrissa primary motor (M1) cortex. The control capabilities of M1 cortex were estimated from experiments with anesthetized animals in which progressive areas along the vibrissa motor branch were microstimulated with rhythmically applied currents. The motion of the vibrissae followed stimulation of M1 cortex only for rates that were well below the frequency of rhythmic whisking; in contrast, the vibrissae followed stimulation

A neural network model for the unsupervised development of sequence-selective units in an artificial neural network is presented. Connection strengths to units develop to encode important non-stationary features of the environment that lie within their initial receptive field. After learning, individual units in the network uniquely encode direction of spectral motion and more general sequential patterns of greater spatial and temporal complexity. Analysis of individual units demonstrates their selective response to subsequences within the learned patterns. I. INTRODUCTION: SEQUENTIAL PATTERNS IN AUDITION While pattern perception is often associated with the categorization of static patterns, the nervous systems of most species must respond to an ongoing barrage of stimulation that, when considered as a neural code, is both temporally and spatially complex. The order in which component stimuli arrive is often an important cue to the identity of the pattern they make up. This is perha...

this paper is to consider some general features of audition in higher animals, human or otherwise, and to describe several simulated components we have developed in our lab that we intend to serve as parts of an auditory system for such open environments. But what should such a system be able to do? Obviously that depends on many details about 2

Sequential neural activity patterns are as ubiquitous as the outputs they drive, which include motor gestures and sequential cognitive processes. Neural sequences are long, compared to the activation durations of participating neurons, and sequence coding is sparse. Numerous studies demonstrate that spike-time-dependent plasticity (STDP), the primary known mechanism for temporal order learning in neurons, cannot organize networks to generate long sequences, raising the question of how such networks are formed. We show that heterosynaptic competition within single neurons, when combined with STDP, organizes networks to generate long unary activity sequences even without sequential training inputs. The network produces a diversity of sequences with a power law length distribution and exponent 1, independent of cellular time constants. We show evidence for a similar distribution of sequence lengths in the recorded premotor song activity of songbirds. These results suggest that neural sequences may be shaped by synaptic constraints and network circuitry rather than cellular time constants.