We present GluonCV and GluonNLP, the deep learning toolkits for computer vision and natural language processing based on Apache MXNet (incubating). These toolkits provide state-of-the-art pre-trained models, training scripts, and training logs, to facilitate rapid prototyping and promote reproducible research. We also provide modular APIs with flexible building blocks to enable efficient customization. Leveraging the MXNet ecosystem, the deep learning models in GluonCV and GluonNLP can be deployed onto a variety of platforms with different programming languages. The Apache 2.0 license has been adopted by GluonCV and GluonNLP to allow for software distribution, modification, and usage.

The success of Deep Learning and its potential use in many safety-critical applicationshas motivated research on formal verification of Neural Network (NN) models. In thiscontext, verification involves proving or disproving that an NN model satisfies certaininput-output properties. Despite the reputation of learned NN models as black boxes,and the theoretical hardness of proving useful properties about them, researchers havebeen successful in verifying some classes of models by exploiting their piecewise linearstructure and taking insights from formal methods such as Satisifiability Modulo Theory.However, these methods are still far from scaling to realistic neural networks. To facilitateprogress on this crucial area, we exploit the Mixed Integer Linear Programming (MIP) formulation of verification to propose a family of algorithms based on Branch-and-Bound (BaB). We show that our family contains previous verification methods as special cases.With the help of the BaB framework, we make three key contributions. Firstly, we identifynew methods that combine the strengths of multiple existing approaches, accomplishingsignificant performance improvements over previous state of the art. Secondly, we introducean effective branching strategy on ReLU non-linearities. This branching strategy allows usto efficiently and successfully deal with high input dimensional problems with convolutionalnetwork architecture, on which previous methods fail frequently. Finally, we proposecomprehensive test data sets and benchmarks which includes a collection of previouslyreleased testcases. We use the data sets to conduct a thorough experimental comparison ofexisting and new algorithms and to provide an inclusive analysis of the factors impactingthe hardness of verification problems.

The transfer learning technique is widely used to learning in one context and applying it to another, i.e. the capacity to apply acquired knowledge and skills to new situations. But is it possible to transfer the learning from a deep neural network to a weaker neural network? Is it possible to improve the performance of a weak neural network using the knowledge acquired by a more powerful neural network? In this work, during the training process of a weak network, we add a loss function that minimizes the distance between the features previously learned from a strong neural network with the features that the weak network must try to learn. To demonstrate the effectiveness and robustness of our approach, we conducted a large number of experiments using three known datasets and demonstrated that a weak neural network can increase its performance if its learning process is driven by a more powerful neural network.

Spatial trajectories are ubiquitous and complex signals. Their analysis is crucial in many research fields, from urban planning to neuroscience. Several approaches have been proposed to cluster trajectories. They rely on hand-crafted features, which struggle to capture the spatio-temporal complexity of the signal, or on Artificial Neural Networks (ANNs) which can be more efficient but less interpretable. In this paper we present a novel ANN architecture designed to capture the spatio-temporal patterns characteristic of a set of trajectories, while taking into account the demographics of the navigators. Hence, our model extracts markers linked to both behaviour and demographics. We propose a composite signal analyser (CompSNN) combining three simple ANN modules. Each of these modules uses different signal representations of the trajectory while remaining interpretable. Our CompSNN performs significantly better than its modules taken in isolation and allows to visualise which parts of the signal were most useful to discriminate the trajectories.

We focus on estimating emph{a priori} generalization error of two-layer ReLU neural networks (NNs) trained by mean squared error, which only depends on initial parameters and the target function, through the following research line. We first estimate emph{a priori} generalization error of finite-width two-layer ReLU NN with constraint of minimal norm solution, which is proved by cite{zhang2019type} to be an equivalent solution of a linearized (w.r.t. parameter) finite-width two-layer NN. As the width goes to infinity, the linearized NN converges to the NN in Neural Tangent Kernel (NTK) regime citep{jacot2018neural}. Thus, we can derive the emph{a priori} generalization error of two-layer ReLU NN in NTK regime. The distance between NN in a NTK regime and a finite-width NN with gradient training is estimated by cite{arora2019exact}. Based on the results in cite{arora2019exact}, our work proves an emph{a priori} generalization error bound of two-layer ReLU NNs. This estimate uses the intrinsic implicit bias of the minimum norm solution without requiring extra regularity in the loss function. This emph{a priori} estimate also implies that NN does not suffer from curse of dimensionality, and a small generalization error can be achieved without requiring exponentially large number of neurons. In addition the research line proposed in this paper can also be used to study other properties of the finite-width network, such as the posterior generalization error.

A deep neural network has relieved the burden of feature engineering by human experts, but comparable efforts are instead required to determine an effective architecture. On the other hands, as the size of a network has over-grown, a lot of resources are also invested to reduce its size. These problems can be addressed by sparsification of an over-complete model, which removes redundant parameters or connections by pruning them away after training or encouraging them to become zero during training. In general, however, these approaches are not fully differentiable and interrupt an end-to-end training process with the stochastic gradient descent in that they require either a parameter selection or a soft-thresholding step. In this paper, we propose a fully differentiable sparsification method for deep neural networks, which allows parameters to be exactly zero during training, and thus can learn the sparsified structure and the weights of networks simultaneously using the stochastic gradient descent. We apply the proposed method to various popular models in order to show its effectiveness.

In classification models fairness can be ensured by solving a constrained optimization problem. We focus on fairness constraints like Disparate Impact, Demographic Parity, and Equalized Odds, which are non-decomposable and non-convex. Researchers define convex surrogates of the constraints and then apply convex optimization frameworks to obtain fair classifiers. Surrogates serve only as an upper bound to the actual constraints, and convexifying fairness constraints might be challenging.

We propose a neural network-based framework, emph{FNNC}, to achieve fairness while maintaining high accuracy in classification. The above fairness constraints are included in the loss using Lagrangian multipliers. We prove bounds on generalization errors for the constrained losses which asymptotically go to zero. The network is optimized using two-step mini-batch stochastic gradient descent. Our experiments show that FNNC performs as good as the state of the art, if not better. The experimental evidence supplements our theoretical guarantees. In summary, we have an automated solution to achieve fairness in classification, which is easily extendable to many fairness constraints.

We introduce an optimized physics-informed neural network (PINN) trained to solve the problem of identifying and characterizing a surface breaking crack in a metal plate. PINNs are neural networks that can combine data and physics in the learning process by adding the residuals of a system of Partial Differential Equations to the loss function. Our PINN is supervised with realistic ultrasonic surface acoustic wave data acquired at a frequency of 5 MHz. The ultrasonic surface wave data is represented as a surface deformation on the top surface of a metal plate, measured by using the method of laser vibrometry. The PINN is physically informed by the acoustic wave equation and its convergence is sped up using adaptive activation functions. The adaptive activation function uses a scalable hyperparameter in the activation function, which is optimized to achieve best performance of the network as it changes dynamically the topology of the loss function involved in the optimization process. The usage of adaptive activation function significantly improves the convergence, notably observed in the current study. We use PINNs to estimate the speed of sound of the metal plate, which we do with an error of 1\%, and then, by allowing the speed of sound to be space dependent, we identify and characterize the crack as the positions where the speed of sound has decreased. Our study also shows the effect of sub-sampling of the data on the sensitivity of sound speed estimates. More broadly, the resulting model shows a promising deep neural network model for ill-posed inverse problems.

As an important type of reinforcement learning algorithms, actor-critic (AC) and natural actor-critic (NAC) algorithms are often executed in two ways for finding optimal policies. In the first nested-loop design, actor's one update of policy is followed by an entire loop of critic's updates of the value function, and the finite-sample analysis of such AC and NAC algorithms have been recently well established. The second two time-scale design, in which actor and critic update simultaneously but with different learning rates, has much fewer tuning parameters than the nested-loop design and is hence substantially easier to implement. Although two time-scale AC and NAC have been shown to converge in the literature, the finite-sample convergence rate has not been established. In this paper, we provide the first such non-asymptotic convergence rate for two time-scale AC and NAC under Markovian sampling and with actor having general policy class approximation. We show that two time-scale AC requires the overall sample complexity at the order of $mathcal{O}(epsilon^{-2.5}log^3(epsilon^{-1}))$ to attain an $epsilon$-accurate stationary point, and two time-scale NAC requires the overall sample complexity at the order of $mathcal{O}(epsilon^{-4}log^2(epsilon^{-1}))$ to attain an $epsilon$-accurate global optimal point. We develop novel techniques for bounding the bias error of the actor due to dynamically changing Markovian sampling and for analyzing the convergence rate of the linear critic with dynamically changing base functions and transition kernel.

Graph Neural Networks (GNNs) are powerful and flexible neural networks that use the naturally sparse connectivity information of the data. GNNs represent this connectivity as sparse matrices, which have lower arithmetic intensity and thus higher communication costs compared to dense matrices, making GNNs harder to scale to high concurrencies than convolutional or fully-connected neural networks.

We present a family of parallel algorithms for training GNNs. These algorithms are based on their counterparts in dense and sparse linear algebra, but they had not been previously applied to GNN training. We show that they can asymptotically reduce communication compared to existing parallel GNN training methods. We implement a promising and practical version that is based on 2D sparse-dense matrix multiplication using torch.distributed. Our implementation parallelizes over GPU-equipped clusters. We train GNNs on up to a hundred GPUs on datasets that include a protein network with over a billion edges.

The notion of incremental learning is to train an ANN algorithm in stages, as and when newer training data arrives. Incremental learning is becoming widespread in recent times with the advent of deep learning. Noise in the training data reduces the accuracy of the algorithm. In this paper, we make an empirical study of the effect of noise in the training phase. We numerically show that the accuracy of the algorithm is dependent more on the location of the error than the percentage of error. Using Perceptron, Feed Forward Neural Network and Radial Basis Function Neural Network, we show that for the same percentage of error, the accuracy of the algorithm significantly varies with the location of error. Furthermore, our results show that the dependence of the accuracy with the location of error is independent of the algorithm. However, the slope of the degradation curve decreases with more sophisticated algorithms

Uncertainties in a structure is inevitable, which generally lead to variation in dynamic response predictions. For a complex structure, brute force Monte Carlo simulation for response variation analysis is infeasible since one single run may already be computationally costly. Data driven meta-modeling approaches have thus been explored to facilitate efficient emulation and statistical inference. The performance of a meta-model hinges upon both the quality and quantity of training dataset. In actual practice, however, high-fidelity data acquired from high-dimensional finite element simulation or experiment are generally scarce, which poses significant challenge to meta-model establishment. In this research, we take advantage of the multi-level response prediction opportunity in structural dynamic analysis, i.e., acquiring rapidly a large amount of low-fidelity data from reduced-order modeling, and acquiring accurately a small amount of high-fidelity data from full-scale finite element analysis. Specifically, we formulate a composite neural network fusion approach that can fully utilize the multi-level, heterogeneous datasets obtained. It implicitly identifies the correlation of the low- and high-fidelity datasets, which yields improved accuracy when compared with the state-of-the-art. Comprehensive investigations using frequency response variation characterization as case example are carried out to demonstrate the performance.

We develop a general framework for data-driven approximation of input-output maps between infinite-dimensional spaces. The proposed approach is motivated by the recent successes of neural networks and deep learning, in combination with ideas from model reduction. This combination results in a neural network approximation which, in principle, is defined on infinite-dimensional spaces and, in practice, is robust to the dimension of finite-dimensional approximations of these spaces required for computation. For a class of input-output maps, and suitably chosen probability measures on the inputs, we prove convergence of the proposed approximation methodology. Numerically we demonstrate the effectiveness of the method on a class of parametric elliptic PDE problems, showing convergence and robustness of the approximation scheme with respect to the size of the discretization, and compare our method with existing algorithms from the literature.

Sparse estimation via penalized likelihood (PL) is now a popular approach to learn the associations among a large set of variables. This paper describes an R package called lslx that implements PL methods for semi-confirmatory structural equation modeling (SEM). In this semi-confirmatory approach, each model parameter can be specified as free/fixed for theory testing, or penalized for exploration. By incorporating either a L1 or minimax concave penalty, the sparsity pattern of the parameter matrix can be efficiently explored. Package lslx minimizes the PL criterion through a quasi-Newton method. The algorithm conducts line search and checks the first-order condition in each iteration to ensure the optimality of the obtained solution. A numerical comparison between competing packages shows that lslx can reliably find PL estimates with the least time. The current package also supports other advanced functionalities, including a two-stage method with auxiliary variables for missing data handling and a reparameterized multi-group SEM to explore population heterogeneity.

Arkaprava Roy, Subhashis Ghosal, Jeffrey Prescott, Kingshuk Roy Choudhury.

Source: The Annals of Applied Statistics, Volume 13, Number 3, 1791--1816.

Abstract:
We study possible relations between Alzheimer’s disease progression and the structure of the connectome which is white matter connecting different regions of the brain. Regression models in covariates including age, gender and disease status for the extent of white matter connecting each pair of regions of the brain are proposed. Subject inhomogeneity is also incorporated in the model through random effects with an unknown distribution. As there is a large number of pairs of regions, we also adopt a dimension reduction technique through graphon ( J. Combin. Theory Ser. B 96 (2006) 933–957) functions which reduces the functions of pairs of regions to functions of regions. The connecting graphon functions are considered unknown but the assumed smoothness allows putting priors of low complexity on these functions. We pursue a nonparametric Bayesian approach by assigning a Dirichlet process scale mixture of zero to mean normal prior on the distributions of the random effects and finite random series of tensor products of B-splines priors on the underlying graphon functions. We develop efficient Markov chain Monte Carlo techniques for drawing samples for the posterior distributions using Hamiltonian Monte Carlo (HMC). The proposed Bayesian method overwhelmingly outperforms a competing method based on ANCOVA models in the simulation setup. The proposed Bayesian approach is applied on a dataset of 100 subjects and 83 brain regions and key regions implicated in the changing connectome are identified.

Jong Hee Park, Yunkyu Sohn.

Source: Bayesian Analysis, Volume 15, Number 1, 133--157.

Abstract:
Dynamic modeling of longitudinal networks has been an increasingly important topic in applied research. While longitudinal network data commonly exhibit dramatic changes in its structures, existing methods have largely focused on modeling smooth topological changes over time. In this paper, we develop a hidden Markov network change-point model (HNC) that combines the multilinear tensor regression model (Hoff, 2011) with a hidden Markov model using Bayesian inference. We model changes in network structure as shifts in discrete states yielding particular sets of network generating parameters. Our simulation results demonstrate that the proposed method correctly detects the number, locations, and types of changes in latent node characteristics. We apply the proposed method to international military alliance networks to find structural changes in the coalition structure among nations.

Nathan B. Cruze, Andreea L. Erciulescu, Balgobin Nandram, Wendy J. Barboza, Linda J. Young.

Source: Statistical Science, Volume 34, Number 2, 301--316.

Abstract:
In the United States, county-level estimates of crop yield, production, and acreage published by the United States Department of Agriculture’s National Agricultural Statistics Service (USDA NASS) play an important role in determining the value of payments allotted to farmers and ranchers enrolled in several federal programs. Given the importance of these official county-level crop estimates, NASS continually strives to improve its crops county estimates program in terms of accuracy, reliability and coverage. In 2015, NASS engaged a panel of experts convened under the auspices of the National Academies of Sciences, Engineering, and Medicine Committee on National Statistics (CNSTAT) for guidance on implementing models that may synthesize multiple sources of information into a single estimate, provide defensible measures of uncertainty, and potentially increase the number of publishable county estimates. The final report titled Improving Crop Estimates by Integrating Multiple Data Sources was released in 2017. This paper discusses several needs and requirements for NASS county-level crop estimates that were illuminated during the activities of the CNSTAT panel. A motivating example of planted acreage estimation in Illinois illustrates several challenges that NASS faces as it considers adopting any explicit model for official crops county estimates.

Jasper Akerboom
Oct 3, 2012; 32:13819-13840
Cellular

Vinod Venkatraman
Mar 9, 2011; 31:3712-3718
BehavioralSystemsCognitive

Brian Zingg
Apr 15, 2020; 40:3250-3267
Systems/Circuits

Amanda Avona
May 29, 2019; 39:4323-4331
Systems/Circuits

Benjamin Meyer
Jul 3, 2019; 39:5326-5335
Systems/Circuits

Earl K. Miller
Aug 15, 1996; 16:5154-5167
Articles

Jasper Akerboom
Oct 3, 2012; 32:13819-13840
Cellular

Thomas Schikorski
Aug 1, 1997; 17:5858-5867
Articles

M Chalfie
Apr 1, 1985; 5:956-964
Articles

You can reduce the story of CRM to a lot of things, especially its many component parts. Social networking, cloud computing and analytics are mentioned often. We don't need an exhaustive list, but if we stop there I think we miss a lot. To me CRM isn't about the parts, although like most people following the industry, I get a modicum of joy when a vendor adds something new to the toolbox.

We can adapt flexibly to environment changes and search for the most appropriate rule to a context. The orbital prefrontal cortex (PFo) has been associated with decision making, rule generation and maintenance, and more generally has been considered important for behavioral flexibility. To better understand the neural mechanisms underlying the flexible behavior, we studied the ability to generate a switching signal in monkey PFo when a strategy is changed. In the strategy task, we used a visual cue to instruct two male rhesus monkeys either to repeat their most recent choice (i.e., stay strategy) or to change it (i.e., shift strategy). To identify the strategy switching-related signal, we compared nonswitch and switch trials, which cued the same or a different strategy from the previous trial, respectively. We found that the switching-related signal emerged during the cue presentation and it was combined with the strategy signal in a subpopulation of cells. Moreover, the error analysis showed that the activity of the switch-related cells reflected whether the monkeys erroneously switched or not the strategy, rather than what was required for that trial. The function of the switching signal could be to prompt the use of different strategies when older strategies are no longer appropriate, conferring the ability to adapt flexibly to environmental changes. In our task, the switching signal might contribute to the implementation of the strategy cued, overcoming potential interference effects from the strategy previously cued. Our results support the idea that ascribes to PFo an important role for behavioral flexibility.

SIGNIFICANCE STATEMENT We can flexibly adapt our behavior to a changing environment. One of the prefrontal areas traditionally associated with the ability to adapt to new contingencies is the orbital prefrontal cortex (PFo). We analyzed the switching related activity using a strategy task in which two rhesus monkeys were instructed by a visual cue either to repeat or change their most recent choice, respectively using a stay or a shift strategy. We found that PFo neurons were modulated by the strategy switching signal, pointing to the importance of PFo in behavioral flexibility by generating control over the switching of strategies.

Hepatocyte growth factor (HGF) is a multifunctional protein that signals through the MET receptor. HGF stimulates cell proliferation, cell dispersion, neuronal survival, and wound healing. In the inner ear, levels of HGF must be fine-tuned for normal hearing. In mice, a deficiency of HGF expression limited to the auditory system, or an overexpression of HGF, causes neurosensory deafness. In humans, noncoding variants in HGF are associated with nonsyndromic deafness DFNB39. However, the mechanism by which these noncoding variants causes deafness was unknown. Here, we reveal the cause of this deafness using a mouse model engineered with a noncoding intronic 10 bp deletion (del10) in Hgf. Male and female mice homozygous for del10 exhibit moderate-to-profound hearing loss at 4 weeks of age as measured by tone burst auditory brainstem responses. The wild type (WT) 80 mV endocochlear potential was significantly reduced in homozygous del10 mice compared with WT littermates. In normal cochlea, endocochlear potentials are dependent on ion homeostasis mediated by the stria vascularis (SV). Previous studies showed that developmental incorporation of neural crest cells into the SV depends on signaling from HGF/MET. We show by immunohistochemistry that, in del10 homozygotes, neural crest cells fail to infiltrate the developing SV intermediate layer. Phenotyping and RNAseq analyses reveal no other significant abnormalities in other tissues. We conclude that, in the inner ear, the noncoding del10 mutation in Hgf leads to developmental defects of the SV and consequently dysfunctional ion homeostasis and a reduction in the EP, recapitulating human DFNB39 nonsyndromic deafness.

SIGNIFICANCE STATEMENT Hereditary deafness is a common, clinically and genetically heterogeneous neurosensory disorder. Previously, we reported that human deafness DFNB39 is associated with noncoding variants in the 3'UTR of a short isoform of HGF encoding hepatocyte growth factor. For normal hearing, HGF levels must be fine-tuned as an excess or deficiency of HGF cause deafness in mouse. Using a Hgf mutant mouse with a small 10 bp deletion recapitulating a human DFNB39 noncoding variant, we demonstrate that neural crest cells fail to migrate into the stria vascularis intermediate layer, resulting in a significantly reduced endocochlear potential, the driving force for sound transduction by inner ear hair cells. HGF-associated deafness is a neurocristopathy but, unlike many other neurocristopathies, it is not syndromic.

It has been proposed that people can generate probabilistic predictions at multiple levels of representation during language comprehension. We used magnetoencephalography (MEG) and electroencephalography (EEG), in combination with representational similarity analysis, to seek neural evidence for the prediction of animacy features. In two studies, MEG and EEG activity was measured as human participants (both sexes) read three-sentence scenarios. Verbs in the final sentences constrained for either animate or inanimate semantic features of upcoming nouns, and the broader discourse context constrained for either a specific noun or for multiple nouns belonging to the same animacy category. We quantified the similarity between spatial patterns of brain activity following the verbs until just before the presentation of the nouns. The MEG and EEG datasets revealed converging evidence that the similarity between spatial patterns of neural activity following animate-constraining verbs was greater than following inanimate-constraining verbs. This effect could not be explained by lexical-semantic processing of the verbs themselves. We therefore suggest that it reflected the inherent difference in the semantic similarity structure of the predicted animate and inanimate nouns. Moreover, the effect was present regardless of whether a specific word could be predicted, providing strong evidence for the prediction of coarse-grained semantic features that goes beyond the prediction of individual words.

SIGNIFICANCE STATEMENT Language inputs unfold very quickly during real-time communication. By predicting ahead, we can give our brains a "head start," so that language comprehension is faster and more efficient. Although most contexts do not constrain strongly for a specific word, they do allow us to predict some upcoming information. For example, following the context of "they cautioned the...," we can predict that the next word will be animate rather than inanimate (we can caution a person, but not an object). Here, we used EEG and MEG techniques to show that the brain is able to use these contextual constraints to predict the animacy of upcoming words during sentence comprehension, and that these predictions are associated with specific spatial patterns of neural activity.

Adaptive coding of stimuli is well documented in perception, where it supports efficient encoding over a broad range of possible percepts. Recently, a similar neural mechanism has been reported also in value-based decision, where it allows optimal encoding of vast ranges of values in PFC: neuronal response to value depends on the choice context (relative coding), rather than being invariant across contexts (absolute coding). Additionally, value learning is sensitive to the amount of feedback information: providing complete feedback (both obtained and forgone outcomes) instead of partial feedback (only obtained outcome) improves learning. However, it is unclear whether relative coding occurs in all PFC regions and how it is affected by feedback information. We systematically investigated univariate and multivariate feedback encoding in various mPFC regions and compared three modes of neural coding: absolute, partially-adaptive and fully-adaptive.

Twenty-eight human participants (both sexes) performed a learning task while undergoing fMRI scanning. On each trial, they chose between two symbols associated with a certain outcome. Then, the decision outcome was revealed. Notably, in one-half of the trials participants received partial feedback, whereas in the other half they got complete feedback. We used univariate and multivariate analysis to explore value encoding in different feedback conditions.

We found that both obtained and forgone outcomes were encoded in mPFC, but with opposite sign in its ventral and dorsal subdivisions. Moreover, we showed that increasing feedback information induced a switch from absolute to relative coding. Our results suggest that complete feedback information enhances context-dependent outcome encoding.

SIGNIFICANCE STATEMENT This study offers a systematic investigation of the effect of the amount of feedback information (partial vs complete) on univariate and multivariate outcome value encoding, within multiple regions in mPFC and cingulate cortex that are critical for value-based decisions and behavioral adaptation. Moreover, we provide the first comparison of three possible models of neural coding (i.e., absolute, partially-adaptive, and fully-adaptive coding) of value signal in these regions, by using commensurable measures of prediction accuracy. Taken together, our results help build a more comprehensive picture of how the human brain encodes and processes outcome value. In particular, our results suggest that simultaneous presentation of obtained and foregone outcomes promotes relative value representation.

Revealing the organization and function of neural circuits is greatly facilitated by viral tools that spread transsynaptically. Adeno-associated virus (AAV) exhibits anterograde transneuronal transport, however, the synaptic specificity of this spread and its broad application within a diverse set of circuits remains to be explored. Here, using anatomic, functional, and molecular approaches, we provide evidence for the preferential transport of AAV1 to postsynaptically connected neurons and reveal its spread is strongly dependent on synaptic transmitter release. In addition to glutamatergic pathways, AAV1 also spreads through GABAergic synapses to both excitatory and inhibitory cell types. We observed little or no transport, however, through neuromodulatory projections (e.g., serotonergic, cholinergic, and noradrenergic). In addition, we found that AAV1 can be transported through long-distance descending projections from various brain regions to effectively transduce spinal cord neurons. Combined with newly designed intersectional and sparse labeling strategies, AAV1 can be applied within a wide variety of pathways to categorize neurons according to their input sources, morphology, and molecular identities. These properties make AAV1 a promising anterograde transsynaptic tool for establishing a comprehensive cell-atlas of the brain, although its capacity for retrograde transport currently limits its use to unidirectional circuits.

SIGNIFICANCE STATEMENT The discovery of anterograde transneuronal spread of AAV1 generates great promise for its application as a unique tool for manipulating input-defined cell populations and mapping their outputs. However, several outstanding questions remain for anterograde transsynaptic approaches in the field: (1) whether AAV1 spreads exclusively or specifically to synaptically connected neurons, and (2) how broad its application could be in various types of neural circuits in the brain. This study provides several lines of evidence in terms of anatomy, functional innervation, and underlying mechanisms, to strongly support that AAV1 anterograde transneuronal spread is highly synapse specific. In addition, several potentially important applications of transsynaptic AAV1 in probing neural circuits are described.

Endogenous neuropeptide Y (NPY) and corticotrophin-releasing factor (CRF) modulate the responses of the basolateral amygdala (BLA) to stress and are associated with the development of stress resilience and vulnerability, respectively. We characterized persistent effects of repeated NPY and CRF treatment on the structure and function of BLA principal neurons in a novel organotypic slice culture (OTC) model of male rat BLA, and examined the contributions of specific NPY receptor subtypes to these neural and behavioral effects. In BLA principal neurons within the OTCs, repeated NPY treatment caused persistent attenuation of excitatory input and induced dendritic hypotrophy via Y₅ receptor activation; conversely, CRF increased excitatory input and induced hypertrophy of BLA principal neurons. Repeated treatment of OTCs with NPY followed by an identical treatment with CRF, or vice versa, inhibited or reversed all structural changes in OTCs. These structural responses to NPY or CRF required calcineurin or CaMKII, respectively. Finally, repeated intra-BLA injections of NPY or a Y₅ receptor agonist increased social interaction, a validated behavior for anxiety, and recapitulated structural changes in BLA neurons seen in OTCs, while a Y₅ receptor antagonist prevented NPY's effects both on behavior and on structure. These results implicate the Y₅ receptor in the long-term, anxiolytic-like effects of NPY in the BLA, consistent with an intrinsic role in stress buffering, and highlight a remarkable mechanism by which BLA neurons may adapt to different levels of stress. Moreover, BLA OTCs offer a robust model to study mechanisms associated with resilience and vulnerability to stress in BLA.

SIGNIFICANCE STATEMENT Within the basolateral amygdala (BLA), neuropeptide Y (NPY) is associated with buffering the neural stress response induced by corticotropin releasing factor, and promoting stress resilience. We used a novel organotypic slice culture model of BLA, complemented with in vivo studies, to examine the cellular mechanisms associated with the actions of NPY. In organotypic slice cultures, repeated NPY treatment reduces the complexity of the dendritic extent of anxiogenic BLA principal neurons, making them less excitable. NPY, via activation of Y5 receptors, additionally inhibits and reverses the increases in dendritic extent and excitability induced by the stress hormone, corticotropin releasing factor. This NPY-mediated neuroplasticity indicates that resilience or vulnerability to stress may thus involve neuropeptide-mediated dendritic remodeling in BLA principal neurons.

Gamma-band oscillations (GBOs) elicited by transient nociceptive stimuli are one of the most promising biomarkers of pain across species. Still, whether these GBOs reflect stimulus encoding in the primary somatosensory cortex (S1) or nocifensive behavior in the primary motor cortex (M1) is debated. Here we recorded neural activity simultaneously from the brain surface as well as at different depths of the bilateral S1/M1 in freely-moving male rats receiving nociceptive stimulation. GBOs measured from superficial layers of S1 contralateral to the stimulated paw not only had the largest magnitude, but also showed the strongest temporal and phase coupling with epidural GBOs. Also, spiking of superficial S1 interneurons had the strongest phase coherence with epidural GBOs. These results provide the first direct demonstration that scalp GBOs, one of the most promising pain biomarkers, reflect neural activity strongly coupled with the fast spiking of interneurons in the superficial layers of the S1 contralateral to the stimulated side.

SIGNIFICANCE STATEMENT Nociceptive-induced gamma-band oscillations (GBOs) measured at population level are one of the most promising biomarkers of pain perception. Our results provide the direct demonstration that these GBOs reflect neural activity coupled with the spike firing of interneurons in the superficial layers of the primary somatosensory cortex (S1) contralateral to the side of nociceptive stimulation. These results address the ongoing debate about whether nociceptive-induced GBOs recorded with scalp EEG or epidurally reflect stimulus encoding in the S1 or nocifensive behavior in the primary motor cortex (M1), and will therefore influence how experiments in pain neuroscience will be designed and interpreted.

We consider the question of how sensory networks enable the detection of sensory stimuli in a combinatorial coding space. We are specifically interested in the olfactory system, wherein recent experimental studies have reported the existence of rich, enigmatic response patterns associated with stimulus onset and offset. This study aims to identify the functional relevance of such response patterns (i.e., what benefits does such neural activity provide in the context of detecting stimuli in a natural environment). We study this problem through the lens of normative, optimization-based modeling. Here, we define the notion of a low-dimensional latent representation of stimulus identity, which is generated through action of the sensory network. The objective of our optimization framework is to ensure high-fidelity tracking of a nominal representation in this latent space in an energy-efficient manner. It turns out that the optimal motifs emerging from this framework possess morphologic similarity with prototypical onset and offset responses observed in vivo in locusts (Schistocerca americana) of either sex. Furthermore, this objective can be exactly achieved by a network with reciprocal excitatory–inhibitory competitive dynamics, similar to interactions between projection neurons and local neurons in the early olfactory system of insects. The derived model also makes several predictions regarding maintenance of robust latent representations in the presence of confounding background information and trade-offs between the energy of sensory activity and resultant behavioral measures such as speed and accuracy of stimulus detection.

SIGNIFICANCE STATEMENT A key area of study in olfactory coding involves understanding the transformation from high-dimensional sensory stimulus to low-dimensional decoded representation. Here, we examine not only the dimensionality reduction of this mapping but also its temporal dynamics, with specific focus on stimuli that are temporally continuous. Through optimization-based synthesis, we examine how sensory networks can track representations without prior assumption of discrete trial structure. We show that such tracking can be achieved by canonical network architectures and dynamics, and that the resulting responses resemble observations from neurons in the insect olfactory system. Thus, our results provide hypotheses regarding the functional role of olfactory circuit activity at both single neuronal and population scales.

Stress alters brain function by modifying the structure and function of neurons and astrocytes. The fine processes of astrocytes are critical for the clearance of neurotransmitters during synaptic transmission. Thus, experience-dependent remodeling of glial processes is anticipated to alter the output of neural circuits. However, the molecular mechanisms that underlie glial structural plasticity are not known. Here we show that a single exposure of male and female mice to an acute stress produced a long-lasting retraction of the lateral processes of cerebellar Bergmann glial cells. These cells express the GluA1 subunit of AMPA-type glutamate receptors, and GluA1 knockdown is known to shorten the length of glial processes. We found that stress reduced the level of GluA1 protein and AMPA receptor-mediated currents in Bergmann glial cells, and these effects were absent in mice devoid of CPEB3, a protein that binds to GluA1 mRNA and regulates GluA1 protein synthesis. Administration of a β-adrenergic receptor blocker attenuated the reduction in GluA1, and deletion of adenylate cyclase 5 prevented GluA1 suppression. Therefore, stress suppresses GluA1 protein synthesis via an adrenergic/adenylyl cyclase/CPEB3 pathway, and reduces the length of astrocyte lateral processes. Our results identify a novel mechanism for GluA1 subunit plasticity in non-neuronal cells and suggest a previously unappreciated role for AMPA receptors in stress-induced astrocytic remodeling.

SIGNIFICANCE STATEMENT Astrocytes play important roles in synaptic transmission by extending fine processes around synapses. In this study, we showed that a single exposure to an acute stress triggered a retraction of lateral/fine processes in mouse cerebellar astrocytes. These astrocytes express GluA1, a glutamate receptor subunit known to lengthen astrocyte processes. We showed that astrocytic structural changes are associated with a reduction of GluA1 protein levels. This requires activation of β-adrenergic receptors and is triggered by noradrenaline released during stress. We identified adenylyl cyclase 5, an enzyme that elevates cAMP levels, as a downstream effector and found that lowering GluA1 levels depends on CPEB3 proteins that bind to GluA1 mRNA. Therefore, stress regulates GluA1 protein synthesis via an adrenergic/adenylyl cyclase/CPEB3 pathway in astrocytes and remodels their fine processes.

As an important cognitive bias, the framing effect shows that our decision preferences are sensitive to the verbal description (i.e., frame) of options. This study focuses on the neural underpinnings of the social framing effect, which is based on decision-making regarding other people. A novel paradigm was used in which participants made a trade-off between economic benefits and the feelings of others. This decision was described as either a "harm" to, or "not helping," other persons in two conditions (Harm frame vs Help frame). Both human males and females were recruited. Participants behaved more prosocially for Harm frame compared with Help frame, resulting in a significant social framing effect. Using functional magnetic resonance imaging, Experiment 1 showed that the social framing effect was associated with stronger activation in the temporoparietal junction (TPJ), especially its right part. The functional connectivity between the right TPJ (rTPJ) and medial prefrontal cortex predicted the social framing effect on the group level. In Experiment 2, we used transcranial direct current stimulation to modulate the activity of the rTPJ and found that the social framing effect became more prominent under anodal (excitatory) stimulation, while the nonsocial framing effect elicited by the economic gain/loss gambling frame remained unaffected. The rTPJ results might be associated with moral conflicts modulated by the social consequences of an action or different levels of mentalizing with others under different frame conditions, but alternative interpretations are also worth noting. These findings could help elucidate the psychological mechanisms of the social framing effect.

SIGNIFICANCE STATEMENT Previous studies have suggested that the framing effect is generated from an interaction between the amygdala and anterior cingulate cortex. This opinion, however, is based on findings from nonsocial framing tasks. Recent research has highlighted the importance of distinguishing between the social and nonsocial framing effects. The current study focuses on the social framing effect and finds out that the temporoparietal junction and its functional connectivity with the medial prefrontal cortex play a significant role. Additionally, modulating the activity of this region leads to changes in social (but not nonsocial) framing effect. Broadly speaking, these findings help understand the difference in neural mechanisms between social and nonsocial decision-making. Meanwhile, they might be illuminating to promote helping behavior in society.

The tuberal hypothalamus is comprised of the dorsomedial, ventromedial, and arcuate nuclei, as well as parts of the lateral hypothalamic area, and it governs a wide range of physiologies. During neurogenesis, tuberal hypothalamic neurons are thought to be born in a dorsal-to-ventral and outside-in pattern, although the accuracy of this description has been questioned over the years. Moreover, the intrinsic factors that control the timing of neurogenesis in this region are poorly characterized. Proneural genes, including Achate-scute-like 1 (Ascl1) and Neurogenin 3 (Neurog3) are widely expressed in hypothalamic progenitors and contribute to lineage commitment and subtype-specific neuronal identifies, but the potential role of Neurogenin 2 (Neurog2) remains unexplored. Birthdating in male and female mice showed that tuberal hypothalamic neurogenesis begins as early as E9.5 in the lateral hypothalamic and arcuate and rapidly expands to dorsomedial and ventromedial neurons by E10.5, peaking throughout the region by E11.5. We confirmed an outside-in trend, except for neurons born at E9.5, and uncovered a rostrocaudal progression but did not confirm a dorsal-ventral patterning to tuberal hypothalamic neuronal birth. In the absence of Neurog2, neurogenesis stalls, with a significant reduction in early-born BrdU⁺ cells but no change at later time points. Further, the loss of Ascl1 yielded a similar delay in neuronal birth, suggesting that Ascl1 cannot rescue the loss of Neurog2 and that these proneural genes act independently in the tuberal hypothalamus. Together, our findings show that Neurog2 functions as a classical proneural gene to regulate the temporal progression of tuberal hypothalamic neurogenesis.

SIGNIFICANCE STATEMENT Here, we investigated the general timing and pattern of neurogenesis within the tuberal hypothalamus. Our results confirmed an outside-in trend of neurogenesis and uncovered a rostrocaudal progression. We also showed that Neurog2 acts as a classical proneural gene and is responsible for regulating the birth of early-born neurons within the ventromedial hypothalamus, acting independently of Ascl1. In addition, we revealed a role for Neurog2 in cell fate specification and differentiation of ventromedial -specific neurons. Last, Neurog2 does not have cross-inhibitory effects on Neurog1, Neurog3, and Ascl1. These findings are the first to reveal a role for Neurog2 in hypothalamic development.

MECP2 gain-of-function and loss-of-function in genetically engineered monkeys recapitulates typical phenotypes in patients with autism, yet where MECP2 mutation affects the monkey brain and whether/how it relates to autism pathology remain unknown. Here we report a combination of gene–circuit–behavior analyses including MECP2 coexpression network, locomotive and cognitive behaviors, and EEG and fMRI findings in 5 MECP2 overexpressed monkeys (Macaca fascicularis; 3 females) and 20 wild-type monkeys (Macaca fascicularis; 11 females). Whole-genome expression analysis revealed MECP2 coexpressed genes significantly enriched in GABA-related signaling pathways, whereby reduced β-synchronization within fronto-parieto-occipital networks was associated with abnormal locomotive behaviors. Meanwhile, MECP2-induced hyperconnectivity in prefrontal and cingulate networks accounted for regressive deficits in reversal learning tasks. Furthermore, we stratified a cohort of 49 patients with autism and 72 healthy controls of 1112 subjects using functional connectivity patterns, and identified dysconnectivity profiles similar to those in monkeys. By establishing a circuit-based construct link between genetically defined models and stratified patients, these results pave new avenues to deconstruct clinical heterogeneity and advance accurate diagnosis in psychiatric disorders.

SIGNIFICANCE STATEMENT Autism spectrum disorder (ASD) is a complex disorder with co-occurring symptoms caused by multiple genetic variations and brain circuit abnormalities. To dissect the gene–circuit–behavior causal chain underlying ASD, animal models are established by manipulating causative genes such as MECP2. However, it is unknown whether such models have captured any circuit-level pathology in ASD patients, as demonstrated by human brain imaging studies. Here, we use transgenic macaques to examine the causal effect of MECP2 overexpression on gene coexpression, brain circuits, and behaviors. For the first time, we demonstrate that the circuit abnormalities linked to MECP2 and autism-like traits in the monkeys can be mapped to a homogeneous ASD subgroup, thereby offering a new strategy to deconstruct clinical heterogeneity in ASD.

Total food losses have been estimated at 1.3 billion tons per year, which represents roughly one-third of the world food production for human consumption. The economic value of food losses and waste amounts to $680 billion in industrialized countries and $310 billion in developing countries. In total, food loss and waste amount to one trillion dollars globally. Lost and wasted food [...]