Richard H. R. Hahnloser

Digital selection and analogue amplification coexist in a cortex-inspiredsilicon circuit

Digital circuits such as the flip-flop use feedback to achieve multi-stability and nonlinearity to restore signals to logical levels, for example 0 and 1. Analogue feedback circuits are generally designed to operate linearly, so that signals are over a range, and the response is unique. By contrast, the response of cortical circuits to sensory stimulation can be both multistable and graded. We propose that the neocortex combines digital selection of an active set of neurons with analogue response by dynamically varying the positive feedback inherent in its recurrent connections. Strong positive feedback causes differential instabilities that drive the selection of a set of active neurons under the constraints embedded in the synaptic weights. Once selected, the active neurons generate weaker, stable feedback that provides analogue amplification of the input. Here we present our model of cortical processing as an electronic circuit that emulates this hybrid operation, and so is able to perform computations that are similar to stimulus selection, gain modulation and spatiotemporal pattern generation in the neocortex.

Neural mechanisms of vocal sequence generation in the songbird

Abstract: Little is known about the biophysical and neuronal circuit mechanisms underlying the generation and learning of behavioral sequences. Songbirds provide a marvelous animal model in which to study these phenomena. By use of a motorized microdrive to record the activity of single neurons in the singing bird, we are beginning to understand the circuits that generate complex vocal sequences. We describe recent experiments elucidating the role of premotor song‐control nucleus HVC in the production of song. We find that HVC neurons projecting to premotor nucleus RA each generate a single burst of spikes at a particular time in the song and may form a sparse representation of temporal order. We incorporate this observation into a working hypothesis for the generation of vocal sequences in the songbird, and examine some implications for song learning.

Spike-time-dependent plasticity and heterosynaptic competition organize networks to produce long scale-free sequences of neural activity.

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.

Neural processing of auditory feedback during vocal practice in a songbird

Songbirds are capable of vocal learning and communication and are ideally suited to the study of neural mechanisms of complex sensory and motor processing. Vocal communication in a noisy bird colony and vocal learning of a specific song template both require the ability to monitor auditory feedback to distinguish self-generated vocalizations from external sounds and to identify mismatches between the developing song and a memorized template acquired from a tutor. However, neurons that respond to auditory feedback from vocal output have not been found in song-control areas despite intensive searching. Here we investigate feedback processing outside the traditional song system, in single auditory forebrain neurons of juvenile zebra finches that were in a late developmental stage of song learning. Overall, we found similarity of spike responses during singing and during playback of the bird’s own song, with song responses commonly leading by a few milliseconds. However, brief time-locked acoustic perturbations of auditory feedback revealed complex sensitivity that could not be predicted from passive playback responses. Some neurons that responded to playback perturbations did not respond to song perturbations, which is reminiscent of sensory-motor mirror neurons. By contrast, some neurons were highly feedback sensitive in that they responded vigorously to song perturbations, but not to unperturbed songs or perturbed playback. These findings suggest that a computational function of forebrain auditory areas may be to detect errors between actual feedback and mirrored feedback deriving from an internal model of the bird’s own song or that of its tutor. Such feedback-sensitive spikes could constitute the key signals that trigger adaptive motor responses to song disruptions or reinforce exploratory motor gestures for vocal learning.

On the piecewise analysis of networks of linear threshold neurons

The computational abilities of recurrent networks of neurons with a linear activation function above threshold are analyzed. These networks selectively realise a linear mapping of their input. Using this property, the dynamics as well as the number and the stability of stationary states can be investigated. The important property of the boundedness of neural activities can be guaranteed by global inhibition. If used together with self-excitation, the global inhibition gives rise to a multi stable winner-take-all (WTA) mechanism. A condition for a neuron to be a potential winner of the competing dynamics is derived. The network becomes a largest input selector when the self-excitation is marginal.Slowing down the global inhibition produces oscillations. The study of oscillations of random networks suggests that all cyclic trajectories of linear threshold networks are a result of the existence of partitions with undamped linear oscillations. Chaotic dynamics were never encountered in computer simulations and perhaps do not exist at all in small networks.

Rapid interhemispheric switching during vocal production in a songbird.

To generate complex bilateral motor patterns such as those underlying birdsong, neural activity must be highly coordinated across the two cerebral hemispheres. However, it remains largely elusive how this coordination is achieved given that interhemispheric communication between song-control areas in the avian cerebrum is restricted to projections received from bilaterally connecting areas in the mid- and hindbrain. By electrically stimulating cerebral premotor areas in zebra finches, we find that behavioral effectiveness of stimulation rapidly switches between hemispheres. In time intervals in which stimulation in one hemisphere tends to distort songs, stimulation in the other hemisphere is mostly ineffective, revealing an idiosyncratic form of motor dominance that bounces back and forth between hemispheres like a virtual ping-pong ball. The intervals of lateralized effectiveness are broadly distributed and are unrelated to simple spectral and temporal song features. Such interhemispheric switching could be an important dynamical aspect of neural coordination that may have evolved from simpler pattern generator circuits.

Feedback interactions between neuronal pointers and maps for attentional processing

Neural networks combining local excitatory feedback with recurrent inhibition are valuable models of neocortical processing. However, incorporating the attentional modulation observed in cortical neurons is problematic. We propose a simple architecture for attentional processing. Our network consists of two reciprocally connected populations of excitatory neurons; a large population (the map) processes a feedforward sensory input, and a small population (the pointer) modulates location and intensity of this processing in an attentional manner dependent on a control input to the pointer. This pointer-map network has rich dynamics despite its simple architecture and explains general computational features related to attention/intention observed in neocortex, making it interesting both theoretically and experimentally.

Silicon synaptic depression.

Abstract. The recent quantitative description of activity-dependent depression in the synaptic transmission between cortical neurons has lead to many interesting suggestions of possible computational implications. Based on a simple biological model, we have constructed an analog circuit that emulates the properties of short-term depressing synapses. The circuit comprises only seven transistors and two capacitors per synapse, and is able to reproduce computational features of depressing synapses such as the 1/F law, the detection of long intervals of presynaptic silence and the sensitivity to redistribution of presynaptic firing rates. It provides a useful basis for implementing neural networks with dynamical synapses.

Auditory representations and memory in birdsong learning

Songbirds are well suited to studies of vocal processing not only because of their impressive motor abilities, but also because of their exquisite sensory system that allows them to detect subtle song variability, memorize complex songs, and monitor auditory feedback during singing. Recent experiments point to areas outside the traditional song system for being relevant to sensory functions implicated in song learning. By manipulating or suppressing activity in these areas, adult birds lose their ability to recognize the songs of their tutors and juveniles are unable to form accurate copies of tutor song. Taken together, these experiments show that the sensory mechanisms for vocal learning encompass a larger network than previously thought.

Spike sorting with hidden Markov models

The ability to detect and sort overlapping spike waveforms in extracellular recordings is key to studies of neural coding at high spatial and temporal resolution. Most spike-sorting algorithms are based on initial spike detection (e.g. by a voltage threshold) and subsequent waveform classification. Much effort has been devoted to the clustering step, despite the fact that conservative spike detection is notoriously difficult in low signal-to-noise conditions and often entails many spike misses. Hidden Markov models (HMMs) can serve as generative models for continuous extracellular data records. These models naturally combine the spike detection and classification steps into a single computational procedure. They unify the advantages of independent component analysis (ICA) and overlap-search algorithms because they blindly perform source separation even in cases where several neurons are recorded on a single electrode. We apply HMMs to artificially generated data and to extracellular signals recorded with glass electrodes. We show that in comparison with state-of-art spike-sorting algorithms, HMM-based spike sorting exhibits a comparable number of false positive spike classifications but many fewer spike misses.