Learning efficient representations of environmental priors in working memory.

Experience shapes our expectations and helps us learn the structure of the environment. Inference models render such learning as a gradual refinement of the observer's estimate of the environmental prior. For instance, when retaining an estimate of an object's features in working memory, learned priors may bias the estimate in the direction of common feature values. Humans display such biases when retaining color estimates on short time intervals. We propose that these systematic biases emerge from modulation of synaptic connectivity in a neural circuit based on the experienced stimulus history, shaping the persistent and collective neural activity that encodes the stimulus estimate. Resulting neural activity attractors are aligned to common stimulus values. Using recently published human response data from a delayed-estimation task in which stimuli (colors) were drawn from a heterogeneous distribution that did not necessarily correspond with reported population biases, we confirm that most subjects' response distributions are better described by experience-dependent learning models than by models with fixed biases. This work suggests systematic limitations in working memory reflect efficient representations of inferred environmental structure, providing new insights into how humans integrate environmental knowledge into their cognitive strategies.

Suboptimal human inference can invert the bias-variance trade-off for decisions with asymmetric evidence.

Solutions to challenging inference problems are often subject to a fundamental trade-off between: 1) bias (being systematically wrong) that is minimized with complex inference strategies, and 2) variance (being oversensitive to uncertain observations) that is minimized with simple inference strategies. However, this trade-off is based on the assumption that the strategies being considered are optimal for their given complexity and thus has unclear relevance to forms of inference based on suboptimal strategies. We examined inference problems applied to rare, asymmetrically available evidence, which a large population of human subjects solved using a diverse set of strategies that varied in form and complexity. In general, subjects using more complex strategies tended to have lower bias and variance, but with a dependence on the form of strategy that reflected an inversion of the classic bias-variance trade-off: subjects who used more complex, but imperfect, Bayesian-like strategies tended to have lower variance but higher bias because of incorrect tuning to latent task features, whereas subjects who used simpler heuristic strategies tended to have higher variance because they operated more directly on the observed samples but lower, near-normative bias. Our results help define new principles that govern individual differences in behavior that depends on rare-event inference and, more generally, about the information-processing trade-offs that can be sensitive to not just the complexity, but also the optimality, of the inference process.

Distinct Excitatory and Inhibitory Bump Wandering in a Stochastic Neural Field.

Localized persistent cortical neural activity is a validated neural substrate of parametric working memory. Such activity "bumps" represent the continuous location of a cue over several seconds. Pyramidal (excitatory (E)) and interneuronal (inhibitory (I)) subpopulations exhibit tuned bumps of activity, linking neural dynamics to behavioral inaccuracies observed in memory recall. However, many bump attractor models collapse these subpopulations into a single joint E/I(lateral inhibitory) population and do not consider the role of interpopulation neural architecture and noise correlations. Both factors have a high potential to impinge upon the stochastic dynamics of these bumps, ultimately shaping behavioral response variance. In our study, we consider a neural field model with separate E/I populations and leverage asymptotic analysis to derive a nonlinear Langevin system describing E/I bump interactions. While the E bump attracts the I bump, the I bump stabilizes but can also repel the E bump, which can result in prolonged relaxation dynamics when both bumps are perturbed. Furthermore, the structure of noise correlations within and between subpopulations strongly shapes the variance in bump position. Surprisingly, higher interpopulation correlations reduce variance.

Ex vivo multi-electrode analysis reveals spatiotemporal dynamics of ictal behavior at the infiltrated margin of glioma.

The purpose of this study is to develop a platform in which the cellular and molecular underpinnings of chronic focal neocortical lesional epilepsy can be explored and use it to characterize seizure-like events (SLEs) in an ex vivo model of infiltrating high-grade glioma. Microelectrode arrays were used to study electrophysiologic changes in ex vivo acute brain slices from a PTEN/p53 deleted, PDGF-B driven mouse model of high-grade glioma. Electrode locations were co-registered to the underlying histology to ascertain the influence of the varying histologic landscape on the observed electrophysiologic changes. Peritumoral, infiltrated, and tumor sites were sampled in tumor-bearing slices. Following the addition of zero Mg2+ solution, all three histologic regions in tumor-bearing slices showed significantly greater increases in firing rates when compared to the control sites. Tumor-bearing slices demonstrated increased proclivity for SLEs, with 40 events in tumor-bearing slices and 5 events in control slices (p-value = .0105). Observed SLEs were characterized by either low voltage fast (LVF) onset patterns or short bursts of repetitive widespread, high amplitude low frequency discharges. Seizure foci comprised areas from all three histologic regions. The onset electrode was found to be at the infiltrated margin in 50% of cases and in the peritumoral region in 36.9% of cases. These findings reveal a landscape of histopathologic and electrophysiologic alterations associated with ictogenesis and spread of tumor-associated seizures.

Cross-scale effects of neural interactions during human neocortical seizure activity.

Small-scale neuronal networks may impose widespread effects on large network dynamics. To unravel this relationship, we analyzed eight multiscale recordings of spontaneous seizures from four patients with epilepsy. During seizures, multiunit spike activity organizes into a submillimeter-sized wavefront, and this activity correlates significantly with low-frequency rhythms from electrocorticographic recordings across a 10-cm-sized neocortical network. Notably, this correlation effect is specific to the ictal wavefront and is absent interictally or from action potential activity outside the wavefront territory. To examine the multiscale interactions, we created a model using a multiscale, nonlinear system and found evidence for a dual role for feedforward inhibition in seizures: while inhibition at the wavefront fails, allowing seizure propagation, feedforward inhibition of the surrounding centimeter-scale networks is activated via long-range excitatory connections. Bifurcation analysis revealed that distinct dynamical pathways for seizure termination depend on the surrounding inhibition strength. Using our model, we found that the mesoscopic, local wavefront acts as the forcing term of the ictal process, while the macroscopic, centimeter-sized network modulates the oscillatory seizure activity.

Role of paroxysmal depolarization in focal seizure activity.

We analyze the role of inhibition in sustaining focal epileptic seizure activity. We review ongoing seizure activity at the mesoscopic scale that can be observed with microelectrode arrays as well as at the macroscale of standard clinical EEG. We provide clinical, experimental, and modeling data to support the hypothesis that paroxysmal depolarization (PD) is a critical component of the ictal machinery. We present dual-patch recordings in cortical cultures showing reduced synaptic transmission associated with presynaptic occurrence of PD, and we find that the PD threshold is cell size related. We further find evidence that optically evoked PD activity in parvalbumin neurons can promote propagation of neuronal excitation in neocortical networks in vitro. Spike sorting results from microelectrode array measurements around ictal wave propagation in human focal seizures demonstrate a strong increase in putative inhibitory firing with an approaching excitatory wave, followed by a sudden reduction of firing at passage. At the macroscopic level, we summarize evidence that this excitatory ictal wave activity is strongly correlated with oscillatory activity across a centimeter-sized cortical network. We summarize Wilson-Cowan-type modeling showing how inhibitory function is crucial for this behavior. Our findings motivated us to develop a network motif of neurons in silico, governed by a reduced version of the Hodgkin-Huxley formalism, to show how feedforward, feedback, PD, and local failure of inhibition contribute to observed dynamics across network scales. The presented multidisciplinary evidence suggests that the PD not only is a cellular marker or epiphenomenon but actively contributes to seizure activity.NEW & NOTEWORTHY We present mechanisms of ongoing focal seizures across meso- and macroscales of microelectrode array and standard clinical recordings, respectively. We find modeling, experimental, and clinical evidence for a dual role of inhibition across these scales: local failure of inhibition allows propagation of a mesoscopic ictal wave, whereas inhibition elsewhere remains intact and sustains macroscopic oscillatory activity. We present evidence for paroxysmal depolarization as a mechanism behind this dual role of inhibition in shaping ictal activity.

Optimal models of decision-making in dynamic environments.

Nature is in constant flux, so animals must account for changes in their environment when making decisions. How animals learn the timescale of such changes and adapt their decision strategies accordingly is not well understood. Recent psychophysical experiments have shown humans and other animals can achieve near-optimal performance at two alternative forced choice (2AFC) tasks in dynamically changing environments. Characterization of performance requires the derivation and analysis of computational models of optimal decision-making policies on such tasks. We review recent theoretical work in this area, and discuss how models compare with subjects' behavior in tasks where the correct choice or evidence quality changes in dynamic, but predictable, ways.

The Relationship Between Ictal Multi-Unit Activity and the Electrocorticogram.

During neocortical seizures in patients with epilepsy, microelectrode array recordings from the ictal core show a strong correlation between the fast, cellular spiking activities and the low-frequency component of the potential field, reflected in the electrocorticogram (ECoG). Here, we model the relationship between the cellular spike activity and this low-frequency component as the input and output signals of a linear time invariant system. Our approach is based on the observation that this relationship can be characterized by a so-called sinc function, the unit impulse response of an ideal (brick-wall) filter. Accordingly, using a brick-wall filter, we are able to convert ictal cellular spike inputs into an output that significantly correlates with the observed seizure activity in the ECoG (r = 0.40 - 0.56,p < 0.01) , while ECoG recordings of subsequent seizures within patients also show significant, but lower, correlations (r = 0.10 - 0.30,p < 0.01) . Furthermore, we can produce seizure-like output signals using synthetic spike trains with ictal properties. We propose a possible physiological mechanism to explain the observed properties associated with an ideal filter, and discuss the potential use of our approach for the evaluation of anticonvulsant strategies.

Multiscale Aspects of Generation of High-Gamma Activity during Seizures in Human Neocortex.

High-gamma (HG; 80-150 Hz) activity in macroscopic clinical records is considered a marker for critical brain regions involved in seizure initiation; it is correlated with pathological multiunit firing during neocortical seizures in the seizure core, an area identified by correlated multiunit spiking and low frequency seizure activity. However, the effects of the spatiotemporal dynamics of seizure on HG power generation are not well understood. Here, we studied HG generation and propagation, using a three-step, multiscale signal analysis and modeling approach. First, we analyzed concurrent neuronal and microscopic network HG activity in neocortical slices from seven intractable epilepsy patients. We found HG activity in these networks, especially when neurons displayed paroxysmal depolarization shifts and network activity was highly synchronized. Second, we examined HG activity acquired with microelectrode arrays recorded during human seizures (n = 8). We confirmed the presence of synchronized HG power across microelectrode records and the macroscale, both specifically associated with the core region of the seizure. Third, we used volume conduction-based modeling to relate HG activity and network synchrony at different network scales. We showed that local HG oscillations require high levels of synchrony to cross scales, and that this requirement is met at the microscopic scale, but not within macroscopic networks. Instead, we present evidence that HG power at the macroscale may result from harmonics of ongoing seizure activity. Ictal HG power marks the seizure core, but the generating mechanism can differ across spatial scales.

Modeling focal epileptic activity in the Wilson-cowan model with depolarization block.

UNLABELLED: Measurements of neuronal signals during human seizure activity and evoked epileptic activity in experimental models suggest that, in these pathological states, the individual nerve cells experience an activity driven depolarization block, i.e. they saturate. We examined the effect of such a saturation in the Wilson-Cowan formalism by adapting the nonlinear activation function; we substituted the commonly applied sigmoid for a Gaussian function. We discuss experimental recordings during a seizure that support this substitution. Next we perform a bifurcation analysis on the Wilson-Cowan model with a Gaussian activation function. The main effect is an additional stable equilibrium with high excitatory and low inhibitory activity. Analysis of coupled local networks then shows that such high activity can stay localized or spread. Specifically, in a spatial continuum we show a wavefront with inhibition leading followed by excitatory activity. We relate our model simulations to observations of spreading activity during seizures. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1186/s13408-015-0019-4) contains supplementary material 1.