Kenneth D. Miller

Competitive Hebbian learning through spike-timing-dependent synaptic plasticity

Hebbian models of development and learning require both activity-dependent synaptic plasticity and a mechanism that induces competition between different synapses. One form of experimentally observed long-term synaptic plasticity, which we call spike-timing-dependent plasticity (STDP), depends on the relative timing of pre- and postsynaptic action potentials. In modeling studies, we find that this form of synaptic modification can automatically balance synaptic strengths to make postsynaptic firing irregular but more sensitive to presynaptic spike timing. It has been argued that neurons in vivo operate in such a balanced regime. Synapses modifiable by STDP compete for control of the timing of postsynaptic action potentials. Inputs that fire the postsynaptic neuron with short latency or that act in correlated groups are able to compete most successfully and develop strong synapses, while synapses of longer-latency or less-effective inputs are weakened.

The role of constraints in Hebbian learning

Models of unsupervised, correlation-based (Hebbian) synaptic plasticity are typically unstable: either all synapses grow until each reaches the maximum allowed strength, or all synapses decay to zero strength. A common method of avoiding these outcomes is to use a constraint that conserves or limits the total synaptic strength over a cell. We study the dynamic effects of such constraints. Two methods of enforcing a constraint are distinguished, multiplicative and subtractive. For otherwise linear learning rules, multiplicative enforcement of a constraint results in dynamics that converge to the principal eigenvector of the operator determining unconstrained synaptic development. Subtractive enforcement, in contrast, typically leads to a final state in which almost all synaptic strengths reach either the maximum or minimum allowed value. This final state is often dominated by weight configurations other than the principal eigenvector of the unconstrained operator. Multiplicative enforcement yields a graded receptive field in which most mutually correlated inputs are represented, whereas subtractive enforcement yields a receptive field that is sharpened to a subset of maximally correlated inputs. If two equivalent input populations (e.g., two eyes) innervate a common target, multiplicative enforcement prevents their segregation (ocular dominance segregation) when the two populations are weakly correlated; whereas subtractive enforcement allows segregation under these circumstances. These results may be used to understand constraints both over output cells and over input cells. A variety of rules that can implement constrained dynamics are discussed.

Adaptive filtering enhances information transmission in visual cortex.

Sensory neuroscience seeks to understand how the brain encodes natural environments. However, neural coding has largely been studied using simplified stimuli. In order to assess whether the brains coding strategy depends on the stimulus ensemble, we apply a new information-theoretic method that allows unbiased calculation of neural filters (receptive fields) from responses to natural scenes or other complex signals with strong multipoint correlations. In the cat primary visual cortex we compare responses to natural inputs with those to noise inputs matched for luminance and contrast. We find that neural filters adaptively change with the input ensemble so as to increase the information carried by the neural response about the filtered stimulus. Adaptation affects the spatial frequency composition of the filter, enhancing sensitivity to under-represented frequencies in agreement with optimal encoding arguments. Adaptation occurs over 40 s to many minutes, longer than most previously reported forms of adaptation.

Physiological gain leads to high ISI variability in a simple model of a cortical regular spiking cell

To understand the interspike interval (ISI) variability displayed by visual cortical neurons (Softky & Koch, 1993), it is critical to examine the dynamics of their neuronal integration, as well as the variability in their synaptic input current. Most previous models have focused on the latter factor. We match a simple integrate-and-fire model to the experimentally measured integrative properties of cortical regular spiking cells (McCormick, Connors, Lighthall, & Prince, 1985). After setting RC parameters, the postspike voltage reset is set to match experimental measurements of neuronal gain (obtained from in vitro plots of firing frequency versus injected current). Examination of the resulting model leads to an intuitive picture of neuronal integration that unifies the seemingly contradictory and random walk pictures that have previously been proposed. When ISIs are dominated by postspike recovery, arguments hold and spiking is regular; after the memory of the last spike becomes negligible, spike threshold crossing is caused by input variance around a steady state and spiking is Poisson. In integrate-and-fire neurons matched to cortical cell physiology, steady-state behavior is predominant, and ISIs are highly variable at all physiological firing rates and for a wide range of inhibitory and excitatory inputs.

Inhibitory Stabilization of the Cortical Network Underlies Visual Surround Suppression

In what regime does the cortical circuit operate? Our intracellular studies of surround suppression in cat primary visual cortex (V1) provide strong evidence on this question. Although suppression has been thought to arise from an increase in lateral inhibition, we find that the inhibition that cells receive is reduced, not increased, by a surround stimulus. Instead, suppression is mediated by a withdrawal of excitation. Thalamic recordings and previous work show that these effects cannot be explained by a withdrawal of thalamic input. We find in theoretical work that this behavior can only arise if V1 operates as an inhibition-stabilized network (ISN), in which excitatory recurrence alone is strong enough to destabilize visual responses but feedback inhibition maintains stability. We confirm two strong tests of this scenario experimentally and show through simulation that observed cell-to-cell variability in surround effects, from facilitation to suppression, can arise naturally from variability in the ISN.

Processing in layer 4 of the neocortical circuit: new insights from visual and somatosensory cortex.

Recent experimental and theoretical results in cat primary visual cortex and in the whisker-barrel fields of rodent primary somatosensory cortex suggest common organizing principles for layer 4, the primary recipient of sensory input from the thalamus. Response tuning of layer 4 cells is largely determined by a local interplay of feed-forward excitation (directly from the thalamus) and inhibition (from layer 4 inhibitory interneurons driven by the thalamus). Feed-forward inhibition dominates excitation, inherits its tuning from the thalamic input, and sharpens the tuning of excitatory cells. Recurrent excitation enhances responses to effective stimuli.

Acoustic variability and distinguishability among mouse ultrasound vocalizations

Auditory neurobiology has benefited significantly from ethological approaches using acoustic communication signals. Developing an ethological model in a genetically manipulable system such as the mouse would enhance the ability to investigate the processing, learning, and recognition of sounds. Characterizing the basic acoustic structure of mouse vocalizations would help lay a foundation for such a future study. Towards this goal, ultrasound vocalizations emitted by isolated mouse pups and pairs of adult males and females have been digitally recorded and examined. Previous work suggests that these calls may have communicative significance. An analysis of the natural variability in their spectral content, median frequency, duration, and repetition period reveals acoustic structure that could be used for recognizing the calls. Other parameters, like the rate of frequency modulation, may also be informative, but have not been examined. Pup isolation calls develop systematically between postnatal day 5 and 12 towards a more stereotyped vocalization--contracting from a wide range of values into narrower clusters of frequency and duration, and shifting from longer to shorter repetition periods. Most significantly, pup isolation and adult encounter calls fall into two distinct spectral and temporal categories, making it possible for a receiver to acoustically distinguish between them, and to potentially categorically perceive them along those dimensions.

One-Dimensional Dynamics of Attention and Decision Making in LIP

Where we allocate our visual spatial attention depends upon a continual competition between internally generated goals and external distractions. Recently it was shown that single neurons in the macaque lateral intraparietal area (LIP) can predict the amount of time a distractor can shift the locus of spatial attention away from a goal. We propose that this remarkable dynamical correspondence between single neurons and attention can be explained by a network model in which generically high-dimensional firing-rate vectors rapidly decay to a single mode. We find direct experimental evidence for this model, not only in the original attentional task, but also in a very different task involving perceptual decision making. These results confirm a theoretical prediction that slowly varying activity patterns are proportional to spontaneous activity, pose constraints on models of persistent activity, and suggest a network mechanism for the emergence of robust behavioral timing from heterogeneous neuronal populations.

IS THE DEVELOPMENT OF ORIENTATION SELECTIVITY INSTRUCTED BY ACTIVITY

Is the development of orientation selectivity in visual cortex instructed by the patterns of neural activity of input neurons? We review evidence as to the role of activity, review models of activity-instructed development, and discuss how these models can be tested. The models can explain the normal development of simple cells with binocularly matched orientation preferences, the effects of monocular deprivation and reverse suture on the orientation map, and the development of a full intracortical circuit sufficient to explain mature response properties including the contrast-invariance of orientation tuning. Existing experiments are consistent with the models, in that (a) selective blockade of ON-center ganglion cells, which will degrade or eliminate the information predicted to drive development of orientation selectivity, in fact prevents development of orientation selectivity; and (b) the spontaneous activities of inputs serving the two eyes are correlated in the lateral geniculate nucleus at appropriate developmental times, as was predicted to be required to achieve binocular matching of preferred orientations. However, definitive tests remain to be done to firmly establish the instructive rather than simply permissive role of activity and determine whether the retinotopically and center type-specific patterns of activity predicted by the models actually exist. We conclude by critically examining alternative scenarios for the development of orientation selectivity and maps, including the idea that maps are genetically prespecified.

Parallel processing by cortical inhibition enables context-dependent behavior

Physical features of sensory stimuli are fixed, but sensory perception is context dependent. The precise mechanisms that govern contextual modulation remain unknown. Here, we trained mice to switch between two contexts: passively listening to pure tones and performing a recognition task for the same stimuli. Two-photon imaging showed that many excitatory neurons in auditory cortex were suppressed during behavior, while some cells became more active. Whole-cell recordings showed that excitatory inputs were affected only modestly by context, but inhibition was more sensitive, with PV+, SOM+, and VIP+ interneurons balancing inhibition and disinhibition within the network. Cholinergic modulation was involved in context switching, with cholinergic axons increasing activity during behavior and directly depolarizing inhibitory cells. Network modeling captured these findings, but only when modulation coincidently drove all three interneuron subtypes, ruling out either inhibition or disinhibition alone as sole mechanism for active engagement. Parallel processing of cholinergic modulation by cortical interneurons therefore enables context-dependent behavior.