The relational bottleneck as an inductive bias for efficient abstraction.

A central challenge for cognitive science is to explain how abstract concepts are acquired from limited experience. This has often been framed in terms of a dichotomy between connectionist and symbolic cognitive models. Here, we highlight a recently emerging line of work that suggests a novel reconciliation of these approaches, by exploiting an inductive bias that we term the relational bottleneck. In that approach, neural networks are constrained via their architecture to focus on relations between perceptual inputs, rather than the attributes of individual inputs. We review a family of models that employ this approach to induce abstractions in a data-efficient manner, emphasizing their potential as candidate models for the acquisition of abstract concepts in the human mind and brain.

A simple, automated method of seizure detection in mouse models of temporal lobe epilepsy.

The lack of preventive and disease modifying therapies for temporal lobe epilepsy (TLE) is a major unmet medical need. Search for such therapies utilize mouse models and require detection of seizures in electroencephalography (EEG) recordings. The labor-intensive nature of reviewing EEGs spanning many weeks underscores the need for a method of automated detection. Here we report a simple automated method of detecting seizures in long term EEG recordings from electrodes implanted in the hippocampus in animal models of TLE. We utilize a 2-pronged approach that relies on the increase in power within the gamma band range (20-50hz) during the seizure followed by suppression of activity following the seizure (post-ictal suppression [PIS]). We demonstrate the utility of this method for detecting seizures in hippocampal and amygdala EEG recordings from multiple models of TLE.

Flow-field inference from neural data using deep recurrent networks.

Computations involved in processes such as decision-making, working memory, and motor control are thought to emerge from the dynamics governing the collective activity of neurons in large populations. But the estimation of these dynamics remains a significant challenge. Here we introduce Flow-field Inference from Neural Data using deep Recurrent networks (FINDR), an unsupervised deep learning method that can infer low-dimensional nonlinear stochastic dynamics underlying neural population activity. Using population spike train data from frontal brain regions of rats performing an auditory decision-making task, we demonstrate that FINDR outperforms existing methods in capturing the heterogeneous responses of individual neurons. We further show that FINDR can discover interpretable low-dimensional dynamics when it is trained to disentangle task-relevant and irrelevant components of the neural population activity. Importantly, the low-dimensional nature of the learned dynamics allows for explicit visualization of flow fields and attractor structures. We suggest FINDR as a powerful method for revealing the low-dimensional task-relevant dynamics of neural populations and their associated computations.

Theory of Gating in Recurrent Neural Networks.

Recurrent neural networks (RNNs) are powerful dynamical models, widely used in machine learning (ML) and neuroscience. Prior theoretical work has focused on RNNs with additive interactions. However gating i.e., multiplicative interactions are ubiquitous in real neurons and also the central feature of the best-performing RNNs in ML. Here, we show that gating offers flexible control of two salient features of the collective dynamics: (i) timescales and (ii) dimensionality. The gate controlling timescales leads to a novel marginally stable state, where the network functions as a flexible integrator. Unlike previous approaches, gating permits this important function without parameter fine-tuning or special symmetries. Gates also provide a flexible, context-dependent mechanism to reset the memory trace, thus complementing the memory function. The gate modulating the dimensionality can induce a novel, discontinuous chaotic transition, where inputs push a stable system to strong chaotic activity, in contrast to the typically stabilizing effect of inputs. At this transition, unlike additive RNNs, the proliferation of critical points (topological complexity) is decoupled from the appearance of chaotic dynamics (dynamical complexity). The rich dynamics are summarized in phase diagrams, thus providing a map for principled parameter initialization choices to ML practitioners.

Disorder and the Neural Representation of Complex Odors.

Animals smelling in the real world use a small number of receptors to sense a vast number of natural molecular mixtures, and proceed to learn arbitrary associations between odors and valences. Here, we propose how the architecture of olfactory circuits leverages disorder, diffuse sensing and redundancy in representation to meet these immense complementary challenges. First, the diffuse and disordered binding of receptors to many molecules compresses a vast but sparsely-structured odor space into a small receptor space, yielding an odor code that preserves similarity in a precise sense. Introducing any order/structure in the sensing degrades similarity preservation. Next, lateral interactions further reduce the correlation present in the low-dimensional receptor code. Finally, expansive disordered projections from the periphery to the central brain reconfigure the densely packed information into a high-dimensional representation, which contains multiple redundant subsets from which downstream neurons can learn flexible associations and valences. Moreover, introducing any order in the expansive projections degrades the ability to recall the learned associations in the presence of noise. We test our theory empirically using data from Drosophila. Our theory suggests that the neural processing of sparse but high-dimensional olfactory information differs from the other senses in its fundamental use of disorder.

ApoE mimetic improves pathology and memory in a model of Alzheimer's disease.

Alzheimer's disease (AD) is the most common form of dementia and is characterized pathologically by Aß plaques. Current treatments are purely symptomatic despite decades of intensive research interest. Notably, patients with the APOE4 allele are at increased risk for developing AD. One hypothesis regarding the mechanism by which the APOE4 allele might increase AD risk is loss of adaptive function, raising the possibility that the exogenous administration of apoE mimetics would have therapeutic effects. In this study, we utilized a previously characterized murine model of AD containing human APP, PS1 and APOE4TR, the APP/PS1/APOETR mouse. We treated male APP/PS1/APOETR mice with the apoE mimetic CN-105 or vehicle for 40d, beginning either at 14-18 or 25-28 weeks of age. After termination of treatment we tested animals in both Morris water maze and contextual fear conditioning, and examined soluble Aß by biochemistery and Aß deposition in cortex by unbiased stereology. We found that transient treatment with CN-105 for 40d beginning at 14-18 weeks reduced Aß pathology and rescued memory deficits in male APP/PS1/APOETR mice. Notably, delaying treatment onset to 25-28 weeks did not produce as robust an effect. These results suggest CN-105 treatment in a mouse model of AD results in a reduction in AD pathology and improved behavioral outcomes when administered early in the course of disease. As CN-105 has an excellent safety profile and is already in clinical trials, these findings raise the possibility that CN-105 represents a novel and translatable therapeutic strategy for AD.

Regression of Epileptogenesis by Inhibiting Tropomyosin Kinase B Signaling following a Seizure.

OBJECTIVE: Temporal lobe epilepsy (TLE) is a devastating disease in which seizures persist in 35% of patients despite optimal use of antiseizure drugs. Clinical and preclinical evidence implicates seizures themselves as one factor promoting epilepsy progression. What is the molecular consequence of a seizure that promotes progression? Evidence from preclinical studies led us to hypothesize that activation of tropomyosin kinase B (TrkB)-phospholipase-C-gamma-1 (PLCγ1) signaling induced by a seizure promotes epileptogenesis. METHODS: To examine the effects of inhibiting TrkB signaling on epileptogenesis following an isolated seizure, we implemented a modified kindling model in which we induced a seizure through amygdala stimulation and then used either a chemical-genetic strategy or pharmacologic methods to disrupt signaling for 2 days following the seizure. The severity of a subsequent seizure was assessed by behavioral and electrographic measures. RESULTS: Transient inhibition of TrkB-PLCγ1 signaling initiated after an isolated seizure limited progression of epileptogenesis, evidenced by the reduced severity and duration of subsequent seizures. Unexpectedly, transient inhibition of TrkB-PLCγ1 signaling initiated following a seizure also reverted a subset of animals to an earlier state of epileptogenesis. Remarkably, inhibition of TrkB-PLCγ1 signaling in the absence of a recent seizure did not reduce severity of subsequent seizures. INTERPRETATION: These results suggest a novel strategy for limiting progression or potentially ameliorating severity of TLE whereby transient inhibition of TrkB-PLCγ1 signaling is initiated following a seizure. ANN NEUROL 2019;86:939-950.

Coding Principles in Adaptation.

Adaptation is a common principle that recurs throughout the nervous system at all stages of processing. This principle manifests in a variety of phenomena, from spike frequency adaptation, to apparent changes in receptive fields with changes in stimulus statistics, to enhanced responses to unexpected stimuli. The ubiquity of adaptation leads naturally to the question: What purpose do these different types of adaptation serve? A diverse set of theories, often highly overlapping, has been proposed to explain the functional role of adaptive phenomena. In this review, we discuss several of these theoretical frameworks, highlighting relationships among them and clarifying distinctions. We summarize observations of the varied manifestations of adaptation, particularly as they relate to these theoretical frameworks, focusing throughout on the visual system and making connections to other sensory systems.

TrkB-Shc Signaling Protects against Hippocampal Injury Following Status Epilepticus.

Temporal lobe epilepsy (TLE) is a common and commonly devastating form of human epilepsy for which only symptomatic therapy is available. One cause of TLE is an episode of de novo prolonged seizures [status epilepticus (SE)]. Understanding the molecular signaling mechanisms by which SE transforms a brain from normal to epileptic may reveal novel targets for preventive and disease-modifying therapies. SE-induced activation of the BDNF receptor tyrosine kinase, TrkB, is one signaling pathway by which SE induces TLE. Although activation of TrkB signaling promotes development of epilepsy in this context, it also reduces SE-induced neuronal death. This led us to hypothesize that distinct signaling pathways downstream of TrkB mediate the desirable (neuroprotective) and undesirable (epileptogenesis) consequences. We subsequently demonstrated that TrkB-mediated activation of phospholipase Cγ1 is required for epileptogenesis. Here we tested the hypothesis that the TrkB-Shc-Akt signaling pathway mediates the neuroprotective consequences of TrkB activation following SE. We studied measures of molecular signaling and cell death in a model of SE in mice of both sexes, including wild-type and TrkBShc/Shc mutant mice in which a point mutation (Y515F) of TrkB prevents the binding of Shc to activated TrkB kinase. Genetic disruption of TrkB-Shc signaling had no effect on severity of SE yet partially inhibited activation of the prosurvival adaptor protein Akt. Importantly, genetic disruption of TrkB-Shc signaling exacerbated hippocampal neuronal death induced by SE. We conclude that therapies targeting TrkB signaling for preventing epilepsy should spare TrkB-Shc-Akt signaling and thereby preserve the neuroprotective benefits.SIGNIFICANCE STATEMENT Temporal lobe epilepsy (TLE) is a common and devastating form of human epilepsy that lacks preventive therapies. Understanding the molecular signaling mechanisms underlying the development of TLE may identify novel therapeutic targets. BDNF signaling thru TrkB receptor tyrosine kinase is one molecular mechanism promoting TLE. We previously discovered that TrkB-mediated activation of phospholipase Cγ1 promotes epileptogenesis. Here we reveal that TrkB-mediated activation of Akt protects against hippocampal neuronal death in vivo following status epilepticus. These findings strengthen the evidence that desirable and undesirable consequences of status epilepticus-induced TrkB activation are mediated by distinct signaling pathways downstream of this receptor. These results provide a strong rationale for a novel therapeutic strategy selectively targeting individual signaling pathways downstream of TrkB for preventing epilepsy.

Arousal-related adjustments of perceptual biases optimize perception in dynamic environments.

Prior expectations can be used to improve perceptual judgments about ambiguous stimuli. However, little is known about if and how these improvements are maintained in dynamic environments in which the quality of appropriate priors changes from one stimulus to the next. Using a sound-localization task, we show that changes in stimulus predictability lead to arousal-mediated adjustments in the magnitude of prior-driven biases that optimize perceptual judgments about each stimulus. These adjustments depend on task-dependent changes in the relevance and reliability of prior expectations, which subjects update using both normative and idiosyncratic principles. The resulting variations in biases across task conditions and individuals are reflected in modulations of pupil diameter, such that larger stimulus-evoked pupil responses correspond to smaller biases. These results suggest a critical role for the arousal system in adjusting the strength of perceptual biases with respect to inferred environmental dynamics to optimize perceptual judgements.

CORRECTED ERROR VIDEO VERSUS A PHYSICAL THERAPIST INSTRUCTED HOME EXERCISE PROGRAM: ACCURACY OF PERFORMING THERAPEUTIC SHOULDER EXERCISES.

BACKGROUND AND PURPOSE: The accurate performance of physical therapy exercises can be difficult. In this evolving healthcare climate it is important to continually look for better methods to educate patients. The use of handouts, in-person demonstration, and video instruction are all potential avenues used to teach proper exercise form. The purpose of this study was to examine if a corrected error video (CEV) would be as effective as a single visit with a physical therapist (PT) to teach healthy subjects how to properly perform four different shoulder rehabilitation exercises. STUDY DESIGN: This was a prospective, single-blinded interventional trial. METHODS: Fifty-eight subjects with no shoulder complaints were recruited from two institutions and randomized into one of two groups: the CEV group (30 subjects) was given a CEV comprised of four shoulder exercises, while the physical therapy group (28 subjects) had one session with a PT as well as a handout of how to complete the exercises. Each subject practiced the exercises for one week and was then videotaped performing them during a return visit. Videos were scored with the shoulder exam assessment tool (SEAT) created by the authors. RESULTS: There was no difference between the groups on total SEAT score (13.66 ± 0.29 vs 13.46 ± 0.30 for CEV vs PT, p = 0.64, 95% CI [-0.06, 0.037]). Average scores for individual exercises also showed no significant difference. CONCLUSION/CLINICAL RELEVANCE: These results demonstrate that the inexpensive and accessible CEV is as beneficial as direct instruction in teaching subjects to properly perform shoulder rehabilitation exercises. LEVEL OF EVIDENCE: 1b.