Computational significance of transient dynamics in cortical networks

Embedding multiple trajectories in simulated recurrent neural networks in a self-organizing manner

Complex neural dynamics produced by the recurrent architecture of neocortical circuits is critical to the cortex's computational power. However, the synaptic learning rules underlying the creation of stable propagation and reproducible neural trajectories within recurrent networks are not understood. Here, we examined synaptic learning rules with the goal of creating recurrent networks in which evoked activity would: (1) propagate throughout the entire network in response to a brief stimulus while avoiding runaway excitation; (2) exhibit spatially and temporally sparse dynamics; and (3) incorporate multiple neural trajectories, i.e., different input patterns should elicit distinct trajectories. We established that an unsupervised learning rule, termed presynaptic-dependent scaling (PSD), can achieve the proposed network dynamics. To quantify the structure of the trained networks, we developed a recurrence index, which revealed that presynaptic-dependent scaling generated a functionally feedforward network when training with a single stimulus. However, training the network with multiple input patterns established that: (1) multiple non-overlapping stable trajectories can be embedded in the network; and (2) the structure of the network became progressively more complex (recurrent) as the number of training patterns increased. In addition, we determined that PSD and spike-timing-dependent plasticity operating in parallel improved the ability of the network to incorporate multiple and less variable trajectories, but also shortened the duration of the neural trajectory. Together, these results establish one of the first learning rules that can embed multiple trajectories, each of which recruits all neurons, within recurrent neural networks in a self-organizing manner.

SORN: a self-organizing recurrent neural network

Understanding the dynamics of recurrent neural networks is crucial for explaining how the brain processes information. In the neocortex, a range of different plasticity mechanisms are shaping recurrent networks into effective information processing circuits that learn appropriate representations for time-varying sensory stimuli. However, it has been difficult to mimic these abilities in artificial neural network models. Here we introduce SORN, a self-organizing recurrent network. It combines three distinct forms of local plasticity to learn spatio-temporal patterns in its input while maintaining its dynamics in a healthy regime suitable for learning. The SORN learns to encode information in the form of trajectories through its high-dimensional state space reminiscent of recent biological findings on cortical coding. All three forms of plasticity are shown to be essential for the network's success.

Recognizing sequences of sequences

The brain's decoding of fast sensory streams is currently impossible to emulate, even approximately, with artificial agents. For example, robust speech recognition is relatively easy for humans but exceptionally difficult for artificial speech recognition systems. In this paper, we propose that recognition can be simplified with an internal model of how sensory input is generated, when formulated in a Bayesian framework. We show that a plausible candidate for an internal or generative model is a hierarchy of ‘stable heteroclinic channels’. This model describes continuous dynamics in the environment as a hierarchy of sequences, where slower sequences cause faster sequences. Under this model, online recognition corresponds to the dynamic decoding of causal sequences, giving a representation of the environment with predictive power on several timescales. We illustrate the ensuing decoding or recognition scheme using synthetic sequences of syllables, where syllables are sequences of phonemes and phonemes are sequences of sound-wave modulations. By presenting anomalous stimuli, we find that the resulting recognition dynamics disclose inference at multiple time scales and are reminiscent of neuronal dynamics seen in the real brain.

Despite tremendous advances in neuroscience, we cannot yet build machines that recognize the world as effortlessly as we do. One reason might be that there are computational approaches to recognition that have not yet been exploited. Here, we demonstrate that the ability to recognize temporal sequences might play an important part. We show that an artificial decoding device can extract natural speech sounds from sound waves if speech is generated as dynamic and transient sequences of sequences. In principle, this means that artificial recognition can be implemented robustly and online using dynamic systems theory and Bayesian inference.

Implications of synaptic biophysics for recurrent network dynamics and active memory

In cortical networks, synaptic excitation is mediated by AMPA- and NMDA-type receptors. NMDA differ from AMPA synaptic potentials with regard to peak current, time course, and a strong voltage-dependent nonlinearity. Here we illustrate based on empirical and computational findings that these specific biophysical properties may have profound implications for the dynamics of cortical networks, and via dynamics on cognitive functions like active memory. The discussion will be led along a minimal set of neural equations introduced to capture the essential dynamics of the various phenomena described. NMDA currents could establish cortical bistability and may provide the relatively constant synaptic drive needed to robustly maintain enhanced levels of activity during working memory epochs, freeing fast AMPA currents for other computational purposes. Perhaps more importantly, variations in NMDA synaptic input-due to their biophysical particularities-control the dynamical regime within which single neurons and networks reside. By provoking bursting, chaotic irregularity, and coherent oscillations their major effect may be on the temporal pattern of spiking activity, rather than on average firing rate. During active memory, neurons may thus be pushed into a spiking regime that harbors complex temporal structure, potentially optimal for the encoding and processing of temporal sequence information. These observations provide a qualitatively different view on the role of synaptic excitation in neocortical dynamics than entailed by many more abstract models. In this sense, this article is a plead for taking the specific biophysics of real neurons and synapses seriously when trying to account for the neurobiology of cognition.

Differential effects of excitatory and inhibitory plasticity on synaptically driven neuronal input-output functions

Ultimately, whether or not a neuron produces a spike determines its contribution to local computations. In response to brief stimuli the probability a neuron will fire can be described by its input-output function, which depends on the net balance and timing of excitatory and inhibitory currents. While excitatory and inhibitory synapses are plastic, most studies examine plasticity of subthreshold events. Thus, the effects of concerted regulation of excitatory and inhibitory synaptic strength on neuronal input-output functions are not well understood. Here, theoretical analyses reveal that excitatory synaptic strength controls the threshold of the neuronal input-output function, while inhibitory plasticity alters the threshold and gain. Experimentally, changes in the balance of excitation and inhibition in CA1 pyramidal neurons also altered their input-output function as predicted by the model. These results support the existence of two functional modes of plasticity that can be used to optimize information processing: threshold and gain plasticity.

Spatial and effector processing in the human parietofrontal network for reaches and saccades

It is generally accepted that interactions between parietal and frontal cortices subserve the visuomotor processing for eye and hand movements. Here, we used a sequential-instruction paradigm in 3-T functional MRI to test the processing of effector and spatial signals, as well as their interaction, as a movement is composed and executed in different stages. Subjects prepared either a saccade or a reach following two successive visual instruction cues, presented in either order. One cue instructed which effector to use (eyes, right hand); the other signaled the spatial goal (leftward vs. rightward target location) of the movement. During the first phase of the prepared movement, after cueing of either goal or effector information, we found significant spatial goal selectivity but no effector specificity along the parietofrontal network. During the second phase of the prepared movement, when both goal and effector information were available, we found a large overlap in the neural circuitry involved in the planning of eye and hand movements. Gradually distributed along this network, we observed clear spatial goal selectivity and limited, but significant, effector specificity. Regions in the intraparietal sulcus and the dorsal premotor cortex were selective to both goal location and motor effector. Taken together, our results suggest that the relative weight of spatial goal and effector selectivity changes along the parietofrontal network, depending on the status of the movement plan.

State-dependent computations: spatiotemporal processing in cortical networks

A conspicuous ability of the brain is to seamlessly assimilate and process spatial and temporal features of sensory stimuli. This ability is indispensable for the recognition of natural stimuli. Yet, a general computational framework for processing spatiotemporal stimuli remains elusive. Recent theoretical and experimental work suggests that spatiotemporal processing emerges from the interaction between incoming stimuli and the internal dynamic state of neural networks, including not only their ongoing spiking activity but also their 'hidden' neuronal states, such as short-term synaptic plasticity.

Fronto-cingulate effective connectivity in major depression: a study with fMRI and dynamic causal modeling

Functional imaging studies are indicating disrupted error monitoring and executive control in a fronto-cingulate network in major depression. However, univariate statistical analyses allow only for a limited assessment of directed neuronal interactions. Therefore, the present study used dynamic causal modeling (DCM) of a fronto-cingulate network to re-analyze the data from a preceding fMRI study in 16 drug-free patients with major depression and 16 healthy controls using the Stroop Color-Word Test (Wagner et al., 2006). In both groups, a significant reciprocal interregional connectivity was found in a cognitive control network including prefrontal cortex (PFC) and dorsal anterior cingulate cortex (ACC). With regard to intrinsic connections we detected a significant difference for dorsal to rostral ACC connectivity between depressive patients and controls in terms of higher connectivity in patients. Additionally, a task by group interaction was observed for the bilinear interaction signaling enhanced task-related input from the dorsal to rostral ACC in subjects with depression. This could be related to the inability of patients to down-regulate rostral ACC activation as observed in the previous univariate analysis. The correlation between interference scores and intrinsic connections from dorsal ACC to dorsolateral PFC (DLPFC) was significant for both groups together, but no significant group differences in correlations could be detected. Thus, the observed relationship between control functions of the dorsal ACC exerted over DLPFC and interference scores appears to be valid in both patients with depression and controls. The findings are consistent with current models of a differential involvement of the fronto-cingulate system in the pathophysiology of major depression.